Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(29)

Side by Side Diff: third_party/sqlite/src/ext/fts3/fts3.c

Issue 901033002: Import SQLite 3.8.7.4. (Closed) Base URL: https://chromium.googlesource.com/chromium/src.git@master
Patch Set: Chromium changes to support SQLite 3.8.7.4. Created 5 years, 10 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
OLDNEW
1 /* 1 /*
2 ** 2006 Oct 10 2 ** 2006 Oct 10
3 ** 3 **
4 ** The author disclaims copyright to this source code. In place of 4 ** The author disclaims copyright to this source code. In place of
5 ** a legal notice, here is a blessing: 5 ** a legal notice, here is a blessing:
6 ** 6 **
7 ** May you do good and not evil. 7 ** May you do good and not evil.
8 ** May you find forgiveness for yourself and forgive others. 8 ** May you find forgiveness for yourself and forgive others.
9 ** May you share freely, never taking more than you give. 9 ** May you share freely, never taking more than you give.
10 ** 10 **
(...skipping 52 matching lines...) Expand 10 before | Expand all | Expand 10 after
63 ** A docid is the unique integer identifier for a single document. 63 ** A docid is the unique integer identifier for a single document.
64 ** A position is the index of a word within the document. The first 64 ** A position is the index of a word within the document. The first
65 ** word of the document has a position of 0. 65 ** word of the document has a position of 0.
66 ** 66 **
67 ** FTS3 used to optionally store character offsets using a compile-time 67 ** FTS3 used to optionally store character offsets using a compile-time
68 ** option. But that functionality is no longer supported. 68 ** option. But that functionality is no longer supported.
69 ** 69 **
70 ** A doclist is stored like this: 70 ** A doclist is stored like this:
71 ** 71 **
72 ** array { 72 ** array {
73 ** varint docid; 73 ** varint docid; (delta from previous doclist)
74 ** array { (position list for column 0) 74 ** array { (position list for column 0)
75 ** varint position; (2 more than the delta from previous position) 75 ** varint position; (2 more than the delta from previous position)
76 ** } 76 ** }
77 ** array { 77 ** array {
78 ** varint POS_COLUMN; (marks start of position list for new column) 78 ** varint POS_COLUMN; (marks start of position list for new column)
79 ** varint column; (index of new column) 79 ** varint column; (index of new column)
80 ** array { 80 ** array {
81 ** varint position; (2 more than the delta from previous position) 81 ** varint position; (2 more than the delta from previous position)
82 ** } 82 ** }
83 ** } 83 ** }
(...skipping 10 matching lines...) Expand all
94 ** 2 for the first position. Example: 94 ** 2 for the first position. Example:
95 ** 95 **
96 ** label: A B C D E F G H I J K 96 ** label: A B C D E F G H I J K
97 ** value: 123 5 9 1 1 14 35 0 234 72 0 97 ** value: 123 5 9 1 1 14 35 0 234 72 0
98 ** 98 **
99 ** The 123 value is the first docid. For column zero in this document 99 ** The 123 value is the first docid. For column zero in this document
100 ** there are two matches at positions 3 and 10 (5-2 and 9-2+3). The 1 100 ** there are two matches at positions 3 and 10 (5-2 and 9-2+3). The 1
101 ** at D signals the start of a new column; the 1 at E indicates that the 101 ** at D signals the start of a new column; the 1 at E indicates that the
102 ** new column is column number 1. There are two positions at 12 and 45 102 ** new column is column number 1. There are two positions at 12 and 45
103 ** (14-2 and 35-2+12). The 0 at H indicate the end-of-document. The 103 ** (14-2 and 35-2+12). The 0 at H indicate the end-of-document. The
104 ** 234 at I is the next docid. It has one position 72 (72-2) and then 104 ** 234 at I is the delta to next docid (357). It has one position 70
105 ** terminates with the 0 at K. 105 ** (72-2) and then terminates with the 0 at K.
106 ** 106 **
107 ** A "position-list" is the list of positions for multiple columns for 107 ** A "position-list" is the list of positions for multiple columns for
108 ** a single docid. A "column-list" is the set of positions for a single 108 ** a single docid. A "column-list" is the set of positions for a single
109 ** column. Hence, a position-list consists of one or more column-lists, 109 ** column. Hence, a position-list consists of one or more column-lists,
110 ** a document record consists of a docid followed by a position-list and 110 ** a document record consists of a docid followed by a position-list and
111 ** a doclist consists of one or more document records. 111 ** a doclist consists of one or more document records.
112 ** 112 **
113 ** A bare doclist omits the position information, becoming an 113 ** A bare doclist omits the position information, becoming an
114 ** array of varint-encoded docids. 114 ** array of varint-encoded docids.
115 ** 115 **
(...skipping 163 matching lines...) Expand 10 before | Expand all | Expand 10 after
279 **** Handling of deletions and updates **** 279 **** Handling of deletions and updates ****
280 ** Since we're using a segmented structure, with no docid-oriented 280 ** Since we're using a segmented structure, with no docid-oriented
281 ** index into the term index, we clearly cannot simply update the term 281 ** index into the term index, we clearly cannot simply update the term
282 ** index when a document is deleted or updated. For deletions, we 282 ** index when a document is deleted or updated. For deletions, we
283 ** write an empty doclist (varint(docid) varint(POS_END)), for updates 283 ** write an empty doclist (varint(docid) varint(POS_END)), for updates
284 ** we simply write the new doclist. Segment merges overwrite older 284 ** we simply write the new doclist. Segment merges overwrite older
285 ** data for a particular docid with newer data, so deletes or updates 285 ** data for a particular docid with newer data, so deletes or updates
286 ** will eventually overtake the earlier data and knock it out. The 286 ** will eventually overtake the earlier data and knock it out. The
287 ** query logic likewise merges doclists so that newer data knocks out 287 ** query logic likewise merges doclists so that newer data knocks out
288 ** older data. 288 ** older data.
289 **
290 ** TODO(shess) Provide a VACUUM type operation to clear out all
291 ** deletions and duplications. This would basically be a forced merge
292 ** into a single segment.
293 */ 289 */
294 #define CHROMIUM_FTS3_CHANGES 1 290 #define CHROMIUM_FTS3_CHANGES 1
295 291
292 #include "fts3Int.h"
296 #if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3) 293 #if !defined(SQLITE_CORE) || defined(SQLITE_ENABLE_FTS3)
297 294
298 #if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE) 295 #if defined(SQLITE_ENABLE_FTS3) && !defined(SQLITE_CORE)
299 # define SQLITE_CORE 1 296 # define SQLITE_CORE 1
300 #endif 297 #endif
301 298
302 #include "fts3Int.h"
303
304 #include <assert.h> 299 #include <assert.h>
305 #include <stdlib.h> 300 #include <stdlib.h>
306 #include <stddef.h> 301 #include <stddef.h>
307 #include <stdio.h> 302 #include <stdio.h>
308 #include <string.h> 303 #include <string.h>
309 #include <stdarg.h> 304 #include <stdarg.h>
310 305
311 #include "fts3.h" 306 #include "fts3.h"
312 #ifndef SQLITE_CORE 307 #ifndef SQLITE_CORE
313 # include "sqlite3ext.h" 308 # include "sqlite3ext.h"
314 SQLITE_EXTENSION_INIT1 309 SQLITE_EXTENSION_INIT1
315 #endif 310 #endif
316 311
312 static int fts3EvalNext(Fts3Cursor *pCsr);
313 static int fts3EvalStart(Fts3Cursor *pCsr);
314 static int fts3TermSegReaderCursor(
315 Fts3Cursor *, const char *, int, int, Fts3MultiSegReader **);
316
317 /* 317 /*
318 ** Write a 64-bit variable-length integer to memory starting at p[0]. 318 ** Write a 64-bit variable-length integer to memory starting at p[0].
319 ** The length of data written will be between 1 and FTS3_VARINT_MAX bytes. 319 ** The length of data written will be between 1 and FTS3_VARINT_MAX bytes.
320 ** The number of bytes written is returned. 320 ** The number of bytes written is returned.
321 */ 321 */
322 int sqlite3Fts3PutVarint(char *p, sqlite_int64 v){ 322 int sqlite3Fts3PutVarint(char *p, sqlite_int64 v){
323 unsigned char *q = (unsigned char *) p; 323 unsigned char *q = (unsigned char *) p;
324 sqlite_uint64 vu = v; 324 sqlite_uint64 vu = v;
325 do{ 325 do{
326 *q++ = (unsigned char) ((vu & 0x7f) | 0x80); 326 *q++ = (unsigned char) ((vu & 0x7f) | 0x80);
327 vu >>= 7; 327 vu >>= 7;
328 }while( vu!=0 ); 328 }while( vu!=0 );
329 q[-1] &= 0x7f; /* turn off high bit in final byte */ 329 q[-1] &= 0x7f; /* turn off high bit in final byte */
330 assert( q - (unsigned char *)p <= FTS3_VARINT_MAX ); 330 assert( q - (unsigned char *)p <= FTS3_VARINT_MAX );
331 return (int) (q - (unsigned char *)p); 331 return (int) (q - (unsigned char *)p);
332 } 332 }
333 333
334 #define GETVARINT_STEP(v, ptr, shift, mask1, mask2, var, ret) \
335 v = (v & mask1) | ( (*ptr++) << shift ); \
336 if( (v & mask2)==0 ){ var = v; return ret; }
337 #define GETVARINT_INIT(v, ptr, shift, mask1, mask2, var, ret) \
338 v = (*ptr++); \
339 if( (v & mask2)==0 ){ var = v; return ret; }
340
334 /* 341 /*
335 ** Read a 64-bit variable-length integer from memory starting at p[0]. 342 ** Read a 64-bit variable-length integer from memory starting at p[0].
336 ** Return the number of bytes read, or 0 on error. 343 ** Return the number of bytes read, or 0 on error.
337 ** The value is stored in *v. 344 ** The value is stored in *v.
338 */ 345 */
339 int sqlite3Fts3GetVarint(const char *p, sqlite_int64 *v){ 346 int sqlite3Fts3GetVarint(const char *p, sqlite_int64 *v){
340 const unsigned char *q = (const unsigned char *) p; 347 const char *pStart = p;
341 sqlite_uint64 x = 0, y = 1; 348 u32 a;
342 while( (*q&0x80)==0x80 && q-(unsigned char *)p<FTS3_VARINT_MAX ){ 349 u64 b;
343 x += y * (*q++ & 0x7f); 350 int shift;
344 y <<= 7; 351
352 GETVARINT_INIT(a, p, 0, 0x00, 0x80, *v, 1);
353 GETVARINT_STEP(a, p, 7, 0x7F, 0x4000, *v, 2);
354 GETVARINT_STEP(a, p, 14, 0x3FFF, 0x200000, *v, 3);
355 GETVARINT_STEP(a, p, 21, 0x1FFFFF, 0x10000000, *v, 4);
356 b = (a & 0x0FFFFFFF );
357
358 for(shift=28; shift<=63; shift+=7){
359 u64 c = *p++;
360 b += (c&0x7F) << shift;
361 if( (c & 0x80)==0 ) break;
345 } 362 }
346 x += y * (*q++); 363 *v = b;
347 *v = (sqlite_int64) x; 364 return (int)(p - pStart);
348 return (int) (q - (unsigned char *)p);
349 } 365 }
350 366
351 /* 367 /*
352 ** Similar to sqlite3Fts3GetVarint(), except that the output is truncated to a 368 ** Similar to sqlite3Fts3GetVarint(), except that the output is truncated to a
353 ** 32-bit integer before it is returned. 369 ** 32-bit integer before it is returned.
354 */ 370 */
355 int sqlite3Fts3GetVarint32(const char *p, int *pi){ 371 int sqlite3Fts3GetVarint32(const char *p, int *pi){
356 sqlite_int64 i; 372 u32 a;
357 int ret = sqlite3Fts3GetVarint(p, &i); 373
358 *pi = (int) i; 374 #ifndef fts3GetVarint32
359 return ret; 375 GETVARINT_INIT(a, p, 0, 0x00, 0x80, *pi, 1);
376 #else
377 a = (*p++);
378 assert( a & 0x80 );
379 #endif
380
381 GETVARINT_STEP(a, p, 7, 0x7F, 0x4000, *pi, 2);
382 GETVARINT_STEP(a, p, 14, 0x3FFF, 0x200000, *pi, 3);
383 GETVARINT_STEP(a, p, 21, 0x1FFFFF, 0x10000000, *pi, 4);
384 a = (a & 0x0FFFFFFF );
385 *pi = (int)(a | ((u32)(*p & 0x0F) << 28));
386 return 5;
360 } 387 }
361 388
362 /* 389 /*
363 ** Return the number of bytes required to encode v as a varint 390 ** Return the number of bytes required to encode v as a varint
364 */ 391 */
365 int sqlite3Fts3VarintLen(sqlite3_uint64 v){ 392 int sqlite3Fts3VarintLen(sqlite3_uint64 v){
366 int i = 0; 393 int i = 0;
367 do{ 394 do{
368 i++; 395 i++;
369 v >>= 7; 396 v >>= 7;
(...skipping 44 matching lines...) Expand 10 before | Expand all | Expand 10 after
414 ** to the first byte past the end of the varint. Add the value of the varint 441 ** to the first byte past the end of the varint. Add the value of the varint
415 ** to *pVal. 442 ** to *pVal.
416 */ 443 */
417 static void fts3GetDeltaVarint(char **pp, sqlite3_int64 *pVal){ 444 static void fts3GetDeltaVarint(char **pp, sqlite3_int64 *pVal){
418 sqlite3_int64 iVal; 445 sqlite3_int64 iVal;
419 *pp += sqlite3Fts3GetVarint(*pp, &iVal); 446 *pp += sqlite3Fts3GetVarint(*pp, &iVal);
420 *pVal += iVal; 447 *pVal += iVal;
421 } 448 }
422 449
423 /* 450 /*
424 ** As long as *pp has not reached its end (pEnd), then do the same 451 ** When this function is called, *pp points to the first byte following a
425 ** as fts3GetDeltaVarint(): read a single varint and add it to *pVal. 452 ** varint that is part of a doclist (or position-list, or any other list
426 ** But if we have reached the end of the varint, just set *pp=0 and 453 ** of varints). This function moves *pp to point to the start of that varint,
427 ** leave *pVal unchanged. 454 ** and sets *pVal by the varint value.
455 **
456 ** Argument pStart points to the first byte of the doclist that the
457 ** varint is part of.
428 */ 458 */
429 static void fts3GetDeltaVarint2(char **pp, char *pEnd, sqlite3_int64 *pVal){ 459 static void fts3GetReverseVarint(
430 if( *pp>=pEnd ){ 460 char **pp,
431 *pp = 0; 461 char *pStart,
432 }else{ 462 sqlite3_int64 *pVal
433 fts3GetDeltaVarint(pp, pVal); 463 ){
434 } 464 sqlite3_int64 iVal;
465 char *p;
466
467 /* Pointer p now points at the first byte past the varint we are
468 ** interested in. So, unless the doclist is corrupt, the 0x80 bit is
469 ** clear on character p[-1]. */
470 for(p = (*pp)-2; p>=pStart && *p&0x80; p--);
471 p++;
472 *pp = p;
473
474 sqlite3Fts3GetVarint(p, &iVal);
475 *pVal = iVal;
435 } 476 }
436 477
437 /* 478 /*
438 ** The xDisconnect() virtual table method. 479 ** The xDisconnect() virtual table method.
439 */ 480 */
440 static int fts3DisconnectMethod(sqlite3_vtab *pVtab){ 481 static int fts3DisconnectMethod(sqlite3_vtab *pVtab){
441 Fts3Table *p = (Fts3Table *)pVtab; 482 Fts3Table *p = (Fts3Table *)pVtab;
442 int i; 483 int i;
443 484
444 assert( p->nPendingData==0 ); 485 assert( p->nPendingData==0 );
445 assert( p->pSegments==0 ); 486 assert( p->pSegments==0 );
446 487
447 /* Free any prepared statements held */ 488 /* Free any prepared statements held */
448 for(i=0; i<SizeofArray(p->aStmt); i++){ 489 for(i=0; i<SizeofArray(p->aStmt); i++){
449 sqlite3_finalize(p->aStmt[i]); 490 sqlite3_finalize(p->aStmt[i]);
450 } 491 }
451 sqlite3_free(p->zSegmentsTbl); 492 sqlite3_free(p->zSegmentsTbl);
452 sqlite3_free(p->zReadExprlist); 493 sqlite3_free(p->zReadExprlist);
453 sqlite3_free(p->zWriteExprlist); 494 sqlite3_free(p->zWriteExprlist);
495 sqlite3_free(p->zContentTbl);
496 sqlite3_free(p->zLanguageid);
454 497
455 /* Invoke the tokenizer destructor to free the tokenizer. */ 498 /* Invoke the tokenizer destructor to free the tokenizer. */
456 p->pTokenizer->pModule->xDestroy(p->pTokenizer); 499 p->pTokenizer->pModule->xDestroy(p->pTokenizer);
457 500
458 sqlite3_free(p); 501 sqlite3_free(p);
459 return SQLITE_OK; 502 return SQLITE_OK;
460 } 503 }
461 504
462 /* 505 /*
463 ** Construct one or more SQL statements from the format string given 506 ** Construct one or more SQL statements from the format string given
(...skipping 19 matching lines...) Expand all
483 }else{ 526 }else{
484 *pRc = sqlite3_exec(db, zSql, 0, 0, 0); 527 *pRc = sqlite3_exec(db, zSql, 0, 0, 0);
485 sqlite3_free(zSql); 528 sqlite3_free(zSql);
486 } 529 }
487 } 530 }
488 531
489 /* 532 /*
490 ** The xDestroy() virtual table method. 533 ** The xDestroy() virtual table method.
491 */ 534 */
492 static int fts3DestroyMethod(sqlite3_vtab *pVtab){ 535 static int fts3DestroyMethod(sqlite3_vtab *pVtab){
536 Fts3Table *p = (Fts3Table *)pVtab;
493 int rc = SQLITE_OK; /* Return code */ 537 int rc = SQLITE_OK; /* Return code */
494 Fts3Table *p = (Fts3Table *)pVtab; 538 const char *zDb = p->zDb; /* Name of database (e.g. "main", "temp") */
495 sqlite3 *db = p->db; 539 sqlite3 *db = p->db; /* Database handle */
496 540
497 /* Drop the shadow tables */ 541 /* Drop the shadow tables */
498 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_content'", p->zDb, p->zName); 542 if( p->zContentTbl==0 ){
499 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_segments'", p->zDb,p->zName); 543 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_content'", zDb, p->zName);
500 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_segdir'", p->zDb, p->zName); 544 }
501 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_docsize'", p->zDb, p->zName); 545 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_segments'", zDb,p->zName);
502 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_stat'", p->zDb, p->zName); 546 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_segdir'", zDb, p->zName);
547 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_docsize'", zDb, p->zName);
548 fts3DbExec(&rc, db, "DROP TABLE IF EXISTS %Q.'%q_stat'", zDb, p->zName);
503 549
504 /* If everything has worked, invoke fts3DisconnectMethod() to free the 550 /* If everything has worked, invoke fts3DisconnectMethod() to free the
505 ** memory associated with the Fts3Table structure and return SQLITE_OK. 551 ** memory associated with the Fts3Table structure and return SQLITE_OK.
506 ** Otherwise, return an SQLite error code. 552 ** Otherwise, return an SQLite error code.
507 */ 553 */
508 return (rc==SQLITE_OK ? fts3DisconnectMethod(pVtab) : rc); 554 return (rc==SQLITE_OK ? fts3DisconnectMethod(pVtab) : rc);
509 } 555 }
510 556
511 557
512 /* 558 /*
513 ** Invoke sqlite3_declare_vtab() to declare the schema for the FTS3 table 559 ** Invoke sqlite3_declare_vtab() to declare the schema for the FTS3 table
514 ** passed as the first argument. This is done as part of the xConnect() 560 ** passed as the first argument. This is done as part of the xConnect()
515 ** and xCreate() methods. 561 ** and xCreate() methods.
516 ** 562 **
517 ** If *pRc is non-zero when this function is called, it is a no-op. 563 ** If *pRc is non-zero when this function is called, it is a no-op.
518 ** Otherwise, if an error occurs, an SQLite error code is stored in *pRc 564 ** Otherwise, if an error occurs, an SQLite error code is stored in *pRc
519 ** before returning. 565 ** before returning.
520 */ 566 */
521 static void fts3DeclareVtab(int *pRc, Fts3Table *p){ 567 static void fts3DeclareVtab(int *pRc, Fts3Table *p){
522 if( *pRc==SQLITE_OK ){ 568 if( *pRc==SQLITE_OK ){
523 int i; /* Iterator variable */ 569 int i; /* Iterator variable */
524 int rc; /* Return code */ 570 int rc; /* Return code */
525 char *zSql; /* SQL statement passed to declare_vtab() */ 571 char *zSql; /* SQL statement passed to declare_vtab() */
526 char *zCols; /* List of user defined columns */ 572 char *zCols; /* List of user defined columns */
573 const char *zLanguageid;
574
575 zLanguageid = (p->zLanguageid ? p->zLanguageid : "__langid");
576 sqlite3_vtab_config(p->db, SQLITE_VTAB_CONSTRAINT_SUPPORT, 1);
527 577
528 /* Create a list of user columns for the virtual table */ 578 /* Create a list of user columns for the virtual table */
529 zCols = sqlite3_mprintf("%Q, ", p->azColumn[0]); 579 zCols = sqlite3_mprintf("%Q, ", p->azColumn[0]);
530 for(i=1; zCols && i<p->nColumn; i++){ 580 for(i=1; zCols && i<p->nColumn; i++){
531 zCols = sqlite3_mprintf("%z%Q, ", zCols, p->azColumn[i]); 581 zCols = sqlite3_mprintf("%z%Q, ", zCols, p->azColumn[i]);
532 } 582 }
533 583
534 /* Create the whole "CREATE TABLE" statement to pass to SQLite */ 584 /* Create the whole "CREATE TABLE" statement to pass to SQLite */
535 zSql = sqlite3_mprintf( 585 zSql = sqlite3_mprintf(
536 "CREATE TABLE x(%s %Q HIDDEN, docid HIDDEN)", zCols, p->zName 586 "CREATE TABLE x(%s %Q HIDDEN, docid HIDDEN, %Q HIDDEN)",
587 zCols, p->zName, zLanguageid
537 ); 588 );
538 if( !zCols || !zSql ){ 589 if( !zCols || !zSql ){
539 rc = SQLITE_NOMEM; 590 rc = SQLITE_NOMEM;
540 }else{ 591 }else{
541 rc = sqlite3_declare_vtab(p->db, zSql); 592 rc = sqlite3_declare_vtab(p->db, zSql);
542 } 593 }
543 594
544 sqlite3_free(zSql); 595 sqlite3_free(zSql);
545 sqlite3_free(zCols); 596 sqlite3_free(zCols);
546 *pRc = rc; 597 *pRc = rc;
547 } 598 }
548 } 599 }
549 600
550 /* 601 /*
602 ** Create the %_stat table if it does not already exist.
603 */
604 void sqlite3Fts3CreateStatTable(int *pRc, Fts3Table *p){
605 fts3DbExec(pRc, p->db,
606 "CREATE TABLE IF NOT EXISTS %Q.'%q_stat'"
607 "(id INTEGER PRIMARY KEY, value BLOB);",
608 p->zDb, p->zName
609 );
610 if( (*pRc)==SQLITE_OK ) p->bHasStat = 1;
611 }
612
613 /*
551 ** Create the backing store tables (%_content, %_segments and %_segdir) 614 ** Create the backing store tables (%_content, %_segments and %_segdir)
552 ** required by the FTS3 table passed as the only argument. This is done 615 ** required by the FTS3 table passed as the only argument. This is done
553 ** as part of the vtab xCreate() method. 616 ** as part of the vtab xCreate() method.
554 ** 617 **
555 ** If the p->bHasDocsize boolean is true (indicating that this is an 618 ** If the p->bHasDocsize boolean is true (indicating that this is an
556 ** FTS4 table, not an FTS3 table) then also create the %_docsize and 619 ** FTS4 table, not an FTS3 table) then also create the %_docsize and
557 ** %_stat tables required by FTS4. 620 ** %_stat tables required by FTS4.
558 */ 621 */
559 static int fts3CreateTables(Fts3Table *p){ 622 static int fts3CreateTables(Fts3Table *p){
560 int rc = SQLITE_OK; /* Return code */ 623 int rc = SQLITE_OK; /* Return code */
561 int i; /* Iterator variable */ 624 int i; /* Iterator variable */
562 char *zContentCols; /* Columns of %_content table */
563 sqlite3 *db = p->db; /* The database connection */ 625 sqlite3 *db = p->db; /* The database connection */
564 626
565 /* Create a list of user columns for the content table */ 627 if( p->zContentTbl==0 ){
566 zContentCols = sqlite3_mprintf("docid INTEGER PRIMARY KEY"); 628 const char *zLanguageid = p->zLanguageid;
567 for(i=0; zContentCols && i<p->nColumn; i++){ 629 char *zContentCols; /* Columns of %_content table */
568 char *z = p->azColumn[i]; 630
569 zContentCols = sqlite3_mprintf("%z, 'c%d%q'", zContentCols, i, z); 631 /* Create a list of user columns for the content table */
632 zContentCols = sqlite3_mprintf("docid INTEGER PRIMARY KEY");
633 for(i=0; zContentCols && i<p->nColumn; i++){
634 char *z = p->azColumn[i];
635 zContentCols = sqlite3_mprintf("%z, 'c%d%q'", zContentCols, i, z);
636 }
637 if( zLanguageid && zContentCols ){
638 zContentCols = sqlite3_mprintf("%z, langid", zContentCols, zLanguageid);
639 }
640 if( zContentCols==0 ) rc = SQLITE_NOMEM;
641
642 /* Create the content table */
643 fts3DbExec(&rc, db,
644 "CREATE TABLE %Q.'%q_content'(%s)",
645 p->zDb, p->zName, zContentCols
646 );
647 sqlite3_free(zContentCols);
570 } 648 }
571 if( zContentCols==0 ) rc = SQLITE_NOMEM;
572 649
573 /* Create the content table */
574 fts3DbExec(&rc, db,
575 "CREATE TABLE %Q.'%q_content'(%s)",
576 p->zDb, p->zName, zContentCols
577 );
578 sqlite3_free(zContentCols);
579 /* Create other tables */ 650 /* Create other tables */
580 fts3DbExec(&rc, db, 651 fts3DbExec(&rc, db,
581 "CREATE TABLE %Q.'%q_segments'(blockid INTEGER PRIMARY KEY, block BLOB);", 652 "CREATE TABLE %Q.'%q_segments'(blockid INTEGER PRIMARY KEY, block BLOB);",
582 p->zDb, p->zName 653 p->zDb, p->zName
583 ); 654 );
584 fts3DbExec(&rc, db, 655 fts3DbExec(&rc, db,
585 "CREATE TABLE %Q.'%q_segdir'(" 656 "CREATE TABLE %Q.'%q_segdir'("
586 "level INTEGER," 657 "level INTEGER,"
587 "idx INTEGER," 658 "idx INTEGER,"
588 "start_block INTEGER," 659 "start_block INTEGER,"
589 "leaves_end_block INTEGER," 660 "leaves_end_block INTEGER,"
590 "end_block INTEGER," 661 "end_block INTEGER,"
591 "root BLOB," 662 "root BLOB,"
592 "PRIMARY KEY(level, idx)" 663 "PRIMARY KEY(level, idx)"
593 ");", 664 ");",
594 p->zDb, p->zName 665 p->zDb, p->zName
595 ); 666 );
596 if( p->bHasDocsize ){ 667 if( p->bHasDocsize ){
597 fts3DbExec(&rc, db, 668 fts3DbExec(&rc, db,
598 "CREATE TABLE %Q.'%q_docsize'(docid INTEGER PRIMARY KEY, size BLOB);", 669 "CREATE TABLE %Q.'%q_docsize'(docid INTEGER PRIMARY KEY, size BLOB);",
599 p->zDb, p->zName 670 p->zDb, p->zName
600 ); 671 );
601 } 672 }
673 assert( p->bHasStat==p->bFts4 );
602 if( p->bHasStat ){ 674 if( p->bHasStat ){
603 fts3DbExec(&rc, db, 675 sqlite3Fts3CreateStatTable(&rc, p);
604 "CREATE TABLE %Q.'%q_stat'(id INTEGER PRIMARY KEY, value BLOB);",
605 p->zDb, p->zName
606 );
607 } 676 }
608 return rc; 677 return rc;
609 } 678 }
610 679
611 /* 680 /*
612 ** Store the current database page-size in bytes in p->nPgsz. 681 ** Store the current database page-size in bytes in p->nPgsz.
613 ** 682 **
614 ** If *pRc is non-zero when this function is called, it is a no-op. 683 ** If *pRc is non-zero when this function is called, it is a no-op.
615 ** Otherwise, if an error occurs, an SQLite error code is stored in *pRc 684 ** Otherwise, if an error occurs, an SQLite error code is stored in *pRc
616 ** before returning. 685 ** before returning.
(...skipping 61 matching lines...) Expand 10 before | Expand all | Expand 10 after
678 int *pRc, /* IN/OUT: Error code */ 747 int *pRc, /* IN/OUT: Error code */
679 char **pz, /* IN/OUT: Pointer to string buffer */ 748 char **pz, /* IN/OUT: Pointer to string buffer */
680 const char *zFormat, /* Printf format string to append */ 749 const char *zFormat, /* Printf format string to append */
681 ... /* Arguments for printf format string */ 750 ... /* Arguments for printf format string */
682 ){ 751 ){
683 if( *pRc==SQLITE_OK ){ 752 if( *pRc==SQLITE_OK ){
684 va_list ap; 753 va_list ap;
685 char *z; 754 char *z;
686 va_start(ap, zFormat); 755 va_start(ap, zFormat);
687 z = sqlite3_vmprintf(zFormat, ap); 756 z = sqlite3_vmprintf(zFormat, ap);
757 va_end(ap);
688 if( z && *pz ){ 758 if( z && *pz ){
689 char *z2 = sqlite3_mprintf("%s%s", *pz, z); 759 char *z2 = sqlite3_mprintf("%s%s", *pz, z);
690 sqlite3_free(z); 760 sqlite3_free(z);
691 z = z2; 761 z = z2;
692 } 762 }
693 if( z==0 ) *pRc = SQLITE_NOMEM; 763 if( z==0 ) *pRc = SQLITE_NOMEM;
694 sqlite3_free(*pz); 764 sqlite3_free(*pz);
695 *pz = z; 765 *pz = z;
696 } 766 }
697 } 767 }
698 768
699 /* 769 /*
700 ** Return a copy of input string zInput enclosed in double-quotes (") and 770 ** Return a copy of input string zInput enclosed in double-quotes (") and
701 ** with all double quote characters escaped. For example: 771 ** with all double quote characters escaped. For example:
702 ** 772 **
703 ** fts3QuoteId("un \"zip\"") -> "un \"\"zip\"\"" 773 ** fts3QuoteId("un \"zip\"") -> "un \"\"zip\"\""
704 ** 774 **
705 ** The pointer returned points to memory obtained from sqlite3_malloc(). It 775 ** The pointer returned points to memory obtained from sqlite3_malloc(). It
706 ** is the callers responsibility to call sqlite3_free() to release this 776 ** is the callers responsibility to call sqlite3_free() to release this
707 ** memory. 777 ** memory.
708 */ 778 */
709 static char *fts3QuoteId(char const *zInput){ 779 static char *fts3QuoteId(char const *zInput){
710 int nRet; 780 int nRet;
711 char *zRet; 781 char *zRet;
712 nRet = 2 + strlen(zInput)*2 + 1; 782 nRet = 2 + (int)strlen(zInput)*2 + 1;
713 zRet = sqlite3_malloc(nRet); 783 zRet = sqlite3_malloc(nRet);
714 if( zRet ){ 784 if( zRet ){
715 int i; 785 int i;
716 char *z = zRet; 786 char *z = zRet;
717 *(z++) = '"'; 787 *(z++) = '"';
718 for(i=0; zInput[i]; i++){ 788 for(i=0; zInput[i]; i++){
719 if( zInput[i]=='"' ) *(z++) = '"'; 789 if( zInput[i]=='"' ) *(z++) = '"';
720 *(z++) = zInput[i]; 790 *(z++) = zInput[i];
721 } 791 }
722 *(z++) = '"'; 792 *(z++) = '"';
723 *(z++) = '\0'; 793 *(z++) = '\0';
724 } 794 }
725 return zRet; 795 return zRet;
726 } 796 }
727 797
728 /* 798 /*
729 ** Return a list of comma separated SQL expressions that could be used 799 ** Return a list of comma separated SQL expressions and a FROM clause that
730 ** in a SELECT statement such as the following: 800 ** could be used in a SELECT statement such as the following:
731 ** 801 **
732 ** SELECT <list of expressions> FROM %_content AS x ... 802 ** SELECT <list of expressions> FROM %_content AS x ...
733 ** 803 **
734 ** to return the docid, followed by each column of text data in order 804 ** to return the docid, followed by each column of text data in order
735 ** from left to write. If parameter zFunc is not NULL, then instead of 805 ** from left to write. If parameter zFunc is not NULL, then instead of
736 ** being returned directly each column of text data is passed to an SQL 806 ** being returned directly each column of text data is passed to an SQL
737 ** function named zFunc first. For example, if zFunc is "unzip" and the 807 ** function named zFunc first. For example, if zFunc is "unzip" and the
738 ** table has the three user-defined columns "a", "b", and "c", the following 808 ** table has the three user-defined columns "a", "b", and "c", the following
739 ** string is returned: 809 ** string is returned:
740 ** 810 **
741 ** "docid, unzip(x.'a'), unzip(x.'b'), unzip(x.'c')" 811 ** "docid, unzip(x.'a'), unzip(x.'b'), unzip(x.'c') FROM %_content AS x"
742 ** 812 **
743 ** The pointer returned points to a buffer allocated by sqlite3_malloc(). It 813 ** The pointer returned points to a buffer allocated by sqlite3_malloc(). It
744 ** is the responsibility of the caller to eventually free it. 814 ** is the responsibility of the caller to eventually free it.
745 ** 815 **
746 ** If *pRc is not SQLITE_OK when this function is called, it is a no-op (and 816 ** If *pRc is not SQLITE_OK when this function is called, it is a no-op (and
747 ** a NULL pointer is returned). Otherwise, if an OOM error is encountered 817 ** a NULL pointer is returned). Otherwise, if an OOM error is encountered
748 ** by this function, NULL is returned and *pRc is set to SQLITE_NOMEM. If 818 ** by this function, NULL is returned and *pRc is set to SQLITE_NOMEM. If
749 ** no error occurs, *pRc is left unmodified. 819 ** no error occurs, *pRc is left unmodified.
750 */ 820 */
751 static char *fts3ReadExprList(Fts3Table *p, const char *zFunc, int *pRc){ 821 static char *fts3ReadExprList(Fts3Table *p, const char *zFunc, int *pRc){
752 char *zRet = 0; 822 char *zRet = 0;
753 char *zFree = 0; 823 char *zFree = 0;
754 char *zFunction; 824 char *zFunction;
755 int i; 825 int i;
756 826
757 if( !zFunc ){ 827 if( p->zContentTbl==0 ){
758 zFunction = ""; 828 if( !zFunc ){
829 zFunction = "";
830 }else{
831 zFree = zFunction = fts3QuoteId(zFunc);
832 }
833 fts3Appendf(pRc, &zRet, "docid");
834 for(i=0; i<p->nColumn; i++){
835 fts3Appendf(pRc, &zRet, ",%s(x.'c%d%q')", zFunction, i, p->azColumn[i]);
836 }
837 if( p->zLanguageid ){
838 fts3Appendf(pRc, &zRet, ", x.%Q", "langid");
839 }
840 sqlite3_free(zFree);
759 }else{ 841 }else{
760 zFree = zFunction = fts3QuoteId(zFunc); 842 fts3Appendf(pRc, &zRet, "rowid");
843 for(i=0; i<p->nColumn; i++){
844 fts3Appendf(pRc, &zRet, ", x.'%q'", p->azColumn[i]);
845 }
846 if( p->zLanguageid ){
847 fts3Appendf(pRc, &zRet, ", x.%Q", p->zLanguageid);
848 }
761 } 849 }
762 fts3Appendf(pRc, &zRet, "docid"); 850 fts3Appendf(pRc, &zRet, " FROM '%q'.'%q%s' AS x",
763 for(i=0; i<p->nColumn; i++){ 851 p->zDb,
764 fts3Appendf(pRc, &zRet, ",%s(x.'c%d%q')", zFunction, i, p->azColumn[i]); 852 (p->zContentTbl ? p->zContentTbl : p->zName),
765 } 853 (p->zContentTbl ? "" : "_content")
766 sqlite3_free(zFree); 854 );
767 return zRet; 855 return zRet;
768 } 856 }
769 857
770 /* 858 /*
771 ** Return a list of N comma separated question marks, where N is the number 859 ** Return a list of N comma separated question marks, where N is the number
772 ** of columns in the %_content table (one for the docid plus one for each 860 ** of columns in the %_content table (one for the docid plus one for each
773 ** user-defined text column). 861 ** user-defined text column).
774 ** 862 **
775 ** If argument zFunc is not NULL, then all but the first question mark 863 ** If argument zFunc is not NULL, then all but the first question mark
776 ** is preceded by zFunc and an open bracket, and followed by a closed 864 ** is preceded by zFunc and an open bracket, and followed by a closed
(...skipping 18 matching lines...) Expand all
795 883
796 if( !zFunc ){ 884 if( !zFunc ){
797 zFunction = ""; 885 zFunction = "";
798 }else{ 886 }else{
799 zFree = zFunction = fts3QuoteId(zFunc); 887 zFree = zFunction = fts3QuoteId(zFunc);
800 } 888 }
801 fts3Appendf(pRc, &zRet, "?"); 889 fts3Appendf(pRc, &zRet, "?");
802 for(i=0; i<p->nColumn; i++){ 890 for(i=0; i<p->nColumn; i++){
803 fts3Appendf(pRc, &zRet, ",%s(?)", zFunction); 891 fts3Appendf(pRc, &zRet, ",%s(?)", zFunction);
804 } 892 }
893 if( p->zLanguageid ){
894 fts3Appendf(pRc, &zRet, ", ?");
895 }
805 sqlite3_free(zFree); 896 sqlite3_free(zFree);
806 return zRet; 897 return zRet;
807 } 898 }
808 899
809 /* 900 /*
901 ** This function interprets the string at (*pp) as a non-negative integer
902 ** value. It reads the integer and sets *pnOut to the value read, then
903 ** sets *pp to point to the byte immediately following the last byte of
904 ** the integer value.
905 **
906 ** Only decimal digits ('0'..'9') may be part of an integer value.
907 **
908 ** If *pp does not being with a decimal digit SQLITE_ERROR is returned and
909 ** the output value undefined. Otherwise SQLITE_OK is returned.
910 **
911 ** This function is used when parsing the "prefix=" FTS4 parameter.
912 */
913 static int fts3GobbleInt(const char **pp, int *pnOut){
914 const char *p; /* Iterator pointer */
915 int nInt = 0; /* Output value */
916
917 for(p=*pp; p[0]>='0' && p[0]<='9'; p++){
918 nInt = nInt * 10 + (p[0] - '0');
919 }
920 if( p==*pp ) return SQLITE_ERROR;
921 *pnOut = nInt;
922 *pp = p;
923 return SQLITE_OK;
924 }
925
926 /*
927 ** This function is called to allocate an array of Fts3Index structures
928 ** representing the indexes maintained by the current FTS table. FTS tables
929 ** always maintain the main "terms" index, but may also maintain one or
930 ** more "prefix" indexes, depending on the value of the "prefix=" parameter
931 ** (if any) specified as part of the CREATE VIRTUAL TABLE statement.
932 **
933 ** Argument zParam is passed the value of the "prefix=" option if one was
934 ** specified, or NULL otherwise.
935 **
936 ** If no error occurs, SQLITE_OK is returned and *apIndex set to point to
937 ** the allocated array. *pnIndex is set to the number of elements in the
938 ** array. If an error does occur, an SQLite error code is returned.
939 **
940 ** Regardless of whether or not an error is returned, it is the responsibility
941 ** of the caller to call sqlite3_free() on the output array to free it.
942 */
943 static int fts3PrefixParameter(
944 const char *zParam, /* ABC in prefix=ABC parameter to parse */
945 int *pnIndex, /* OUT: size of *apIndex[] array */
946 struct Fts3Index **apIndex /* OUT: Array of indexes for this table */
947 ){
948 struct Fts3Index *aIndex; /* Allocated array */
949 int nIndex = 1; /* Number of entries in array */
950
951 if( zParam && zParam[0] ){
952 const char *p;
953 nIndex++;
954 for(p=zParam; *p; p++){
955 if( *p==',' ) nIndex++;
956 }
957 }
958
959 aIndex = sqlite3_malloc(sizeof(struct Fts3Index) * nIndex);
960 *apIndex = aIndex;
961 *pnIndex = nIndex;
962 if( !aIndex ){
963 return SQLITE_NOMEM;
964 }
965
966 memset(aIndex, 0, sizeof(struct Fts3Index) * nIndex);
967 if( zParam ){
968 const char *p = zParam;
969 int i;
970 for(i=1; i<nIndex; i++){
971 int nPrefix;
972 if( fts3GobbleInt(&p, &nPrefix) ) return SQLITE_ERROR;
973 aIndex[i].nPrefix = nPrefix;
974 p++;
975 }
976 }
977
978 return SQLITE_OK;
979 }
980
981 /*
982 ** This function is called when initializing an FTS4 table that uses the
983 ** content=xxx option. It determines the number of and names of the columns
984 ** of the new FTS4 table.
985 **
986 ** The third argument passed to this function is the value passed to the
987 ** config=xxx option (i.e. "xxx"). This function queries the database for
988 ** a table of that name. If found, the output variables are populated
989 ** as follows:
990 **
991 ** *pnCol: Set to the number of columns table xxx has,
992 **
993 ** *pnStr: Set to the total amount of space required to store a copy
994 ** of each columns name, including the nul-terminator.
995 **
996 ** *pazCol: Set to point to an array of *pnCol strings. Each string is
997 ** the name of the corresponding column in table xxx. The array
998 ** and its contents are allocated using a single allocation. It
999 ** is the responsibility of the caller to free this allocation
1000 ** by eventually passing the *pazCol value to sqlite3_free().
1001 **
1002 ** If the table cannot be found, an error code is returned and the output
1003 ** variables are undefined. Or, if an OOM is encountered, SQLITE_NOMEM is
1004 ** returned (and the output variables are undefined).
1005 */
1006 static int fts3ContentColumns(
1007 sqlite3 *db, /* Database handle */
1008 const char *zDb, /* Name of db (i.e. "main", "temp" etc.) */
1009 const char *zTbl, /* Name of content table */
1010 const char ***pazCol, /* OUT: Malloc'd array of column names */
1011 int *pnCol, /* OUT: Size of array *pazCol */
1012 int *pnStr /* OUT: Bytes of string content */
1013 ){
1014 int rc = SQLITE_OK; /* Return code */
1015 char *zSql; /* "SELECT *" statement on zTbl */
1016 sqlite3_stmt *pStmt = 0; /* Compiled version of zSql */
1017
1018 zSql = sqlite3_mprintf("SELECT * FROM %Q.%Q", zDb, zTbl);
1019 if( !zSql ){
1020 rc = SQLITE_NOMEM;
1021 }else{
1022 rc = sqlite3_prepare(db, zSql, -1, &pStmt, 0);
1023 }
1024 sqlite3_free(zSql);
1025
1026 if( rc==SQLITE_OK ){
1027 const char **azCol; /* Output array */
1028 int nStr = 0; /* Size of all column names (incl. 0x00) */
1029 int nCol; /* Number of table columns */
1030 int i; /* Used to iterate through columns */
1031
1032 /* Loop through the returned columns. Set nStr to the number of bytes of
1033 ** space required to store a copy of each column name, including the
1034 ** nul-terminator byte. */
1035 nCol = sqlite3_column_count(pStmt);
1036 for(i=0; i<nCol; i++){
1037 const char *zCol = sqlite3_column_name(pStmt, i);
1038 nStr += (int)strlen(zCol) + 1;
1039 }
1040
1041 /* Allocate and populate the array to return. */
1042 azCol = (const char **)sqlite3_malloc(sizeof(char *) * nCol + nStr);
1043 if( azCol==0 ){
1044 rc = SQLITE_NOMEM;
1045 }else{
1046 char *p = (char *)&azCol[nCol];
1047 for(i=0; i<nCol; i++){
1048 const char *zCol = sqlite3_column_name(pStmt, i);
1049 int n = (int)strlen(zCol)+1;
1050 memcpy(p, zCol, n);
1051 azCol[i] = p;
1052 p += n;
1053 }
1054 }
1055 sqlite3_finalize(pStmt);
1056
1057 /* Set the output variables. */
1058 *pnCol = nCol;
1059 *pnStr = nStr;
1060 *pazCol = azCol;
1061 }
1062
1063 return rc;
1064 }
1065
1066 /*
810 ** This function is the implementation of both the xConnect and xCreate 1067 ** This function is the implementation of both the xConnect and xCreate
811 ** methods of the FTS3 virtual table. 1068 ** methods of the FTS3 virtual table.
812 ** 1069 **
813 ** The argv[] array contains the following: 1070 ** The argv[] array contains the following:
814 ** 1071 **
815 ** argv[0] -> module name ("fts3" or "fts4") 1072 ** argv[0] -> module name ("fts3" or "fts4")
816 ** argv[1] -> database name 1073 ** argv[1] -> database name
817 ** argv[2] -> table name 1074 ** argv[2] -> table name
818 ** argv[...] -> "column name" and other module argument fields. 1075 ** argv[...] -> "column name" and other module argument fields.
819 */ 1076 */
(...skipping 11 matching lines...) Expand all
831 int rc = SQLITE_OK; /* Return code */ 1088 int rc = SQLITE_OK; /* Return code */
832 int i; /* Iterator variable */ 1089 int i; /* Iterator variable */
833 int nByte; /* Size of allocation used for *p */ 1090 int nByte; /* Size of allocation used for *p */
834 int iCol; /* Column index */ 1091 int iCol; /* Column index */
835 int nString = 0; /* Bytes required to hold all column names */ 1092 int nString = 0; /* Bytes required to hold all column names */
836 int nCol = 0; /* Number of columns in the FTS table */ 1093 int nCol = 0; /* Number of columns in the FTS table */
837 char *zCsr; /* Space for holding column names */ 1094 char *zCsr; /* Space for holding column names */
838 int nDb; /* Bytes required to hold database name */ 1095 int nDb; /* Bytes required to hold database name */
839 int nName; /* Bytes required to hold table name */ 1096 int nName; /* Bytes required to hold table name */
840 int isFts4 = (argv[0][3]=='4'); /* True for FTS4, false for FTS3 */ 1097 int isFts4 = (argv[0][3]=='4'); /* True for FTS4, false for FTS3 */
841 int bNoDocsize = 0; /* True to omit %_docsize table */
842 const char **aCol; /* Array of column names */ 1098 const char **aCol; /* Array of column names */
843 sqlite3_tokenizer *pTokenizer = 0; /* Tokenizer for this table */ 1099 sqlite3_tokenizer *pTokenizer = 0; /* Tokenizer for this table */
844 1100
845 char *zCompress = 0; 1101 int nIndex; /* Size of aIndex[] array */
846 char *zUncompress = 0; 1102 struct Fts3Index *aIndex = 0; /* Array of indexes for this table */
1103
1104 /* The results of parsing supported FTS4 key=value options: */
1105 int bNoDocsize = 0; /* True to omit %_docsize table */
1106 int bDescIdx = 0; /* True to store descending indexes */
1107 char *zPrefix = 0; /* Prefix parameter value (or NULL) */
1108 char *zCompress = 0; /* compress=? parameter (or NULL) */
1109 char *zUncompress = 0; /* uncompress=? parameter (or NULL) */
1110 char *zContent = 0; /* content=? parameter (or NULL) */
1111 char *zLanguageid = 0; /* languageid=? parameter (or NULL) */
1112 char **azNotindexed = 0; /* The set of notindexed= columns */
1113 int nNotindexed = 0; /* Size of azNotindexed[] array */
847 1114
848 assert( strlen(argv[0])==4 ); 1115 assert( strlen(argv[0])==4 );
849 assert( (sqlite3_strnicmp(argv[0], "fts4", 4)==0 && isFts4) 1116 assert( (sqlite3_strnicmp(argv[0], "fts4", 4)==0 && isFts4)
850 || (sqlite3_strnicmp(argv[0], "fts3", 4)==0 && !isFts4) 1117 || (sqlite3_strnicmp(argv[0], "fts3", 4)==0 && !isFts4)
851 ); 1118 );
852 1119
853 nDb = (int)strlen(argv[1]) + 1; 1120 nDb = (int)strlen(argv[1]) + 1;
854 nName = (int)strlen(argv[2]) + 1; 1121 nName = (int)strlen(argv[2]) + 1;
855 1122
856 aCol = (const char **)sqlite3_malloc(sizeof(const char *) * (argc-2) ); 1123 nByte = sizeof(const char *) * (argc-2);
857 if( !aCol ) return SQLITE_NOMEM; 1124 aCol = (const char **)sqlite3_malloc(nByte);
858 memset((void *)aCol, 0, sizeof(const char *) * (argc-2)); 1125 if( aCol ){
1126 memset((void*)aCol, 0, nByte);
1127 azNotindexed = (char **)sqlite3_malloc(nByte);
1128 }
1129 if( azNotindexed ){
1130 memset(azNotindexed, 0, nByte);
1131 }
1132 if( !aCol || !azNotindexed ){
1133 rc = SQLITE_NOMEM;
1134 goto fts3_init_out;
1135 }
859 1136
860 /* Loop through all of the arguments passed by the user to the FTS3/4 1137 /* Loop through all of the arguments passed by the user to the FTS3/4
861 ** module (i.e. all the column names and special arguments). This loop 1138 ** module (i.e. all the column names and special arguments). This loop
862 ** does the following: 1139 ** does the following:
863 ** 1140 **
864 ** + Figures out the number of columns the FTSX table will have, and 1141 ** + Figures out the number of columns the FTSX table will have, and
865 ** the number of bytes of space that must be allocated to store copies 1142 ** the number of bytes of space that must be allocated to store copies
866 ** of the column names. 1143 ** of the column names.
867 ** 1144 **
868 ** + If there is a tokenizer specification included in the arguments, 1145 ** + If there is a tokenizer specification included in the arguments,
869 ** initializes the tokenizer pTokenizer. 1146 ** initializes the tokenizer pTokenizer.
870 */ 1147 */
871 for(i=3; rc==SQLITE_OK && i<argc; i++){ 1148 for(i=3; rc==SQLITE_OK && i<argc; i++){
872 char const *z = argv[i]; 1149 char const *z = argv[i];
873 int nKey; 1150 int nKey;
874 char *zVal; 1151 char *zVal;
875 1152
876 /* Check if this is a tokenizer specification */ 1153 /* Check if this is a tokenizer specification */
877 if( !pTokenizer 1154 if( !pTokenizer
878 && strlen(z)>8 1155 && strlen(z)>8
879 && 0==sqlite3_strnicmp(z, "tokenize", 8) 1156 && 0==sqlite3_strnicmp(z, "tokenize", 8)
880 && 0==sqlite3Fts3IsIdChar(z[8]) 1157 && 0==sqlite3Fts3IsIdChar(z[8])
881 ){ 1158 ){
882 rc = sqlite3Fts3InitTokenizer(pHash, &z[9], &pTokenizer, pzErr); 1159 rc = sqlite3Fts3InitTokenizer(pHash, &z[9], &pTokenizer, pzErr);
883 } 1160 }
884 1161
885 /* Check if it is an FTS4 special argument. */ 1162 /* Check if it is an FTS4 special argument. */
886 else if( isFts4 && fts3IsSpecialColumn(z, &nKey, &zVal) ){ 1163 else if( isFts4 && fts3IsSpecialColumn(z, &nKey, &zVal) ){
1164 struct Fts4Option {
1165 const char *zOpt;
1166 int nOpt;
1167 } aFts4Opt[] = {
1168 { "matchinfo", 9 }, /* 0 -> MATCHINFO */
1169 { "prefix", 6 }, /* 1 -> PREFIX */
1170 { "compress", 8 }, /* 2 -> COMPRESS */
1171 { "uncompress", 10 }, /* 3 -> UNCOMPRESS */
1172 { "order", 5 }, /* 4 -> ORDER */
1173 { "content", 7 }, /* 5 -> CONTENT */
1174 { "languageid", 10 }, /* 6 -> LANGUAGEID */
1175 { "notindexed", 10 } /* 7 -> NOTINDEXED */
1176 };
1177
1178 int iOpt;
887 if( !zVal ){ 1179 if( !zVal ){
888 rc = SQLITE_NOMEM; 1180 rc = SQLITE_NOMEM;
889 goto fts3_init_out; 1181 }else{
1182 for(iOpt=0; iOpt<SizeofArray(aFts4Opt); iOpt++){
1183 struct Fts4Option *pOp = &aFts4Opt[iOpt];
1184 if( nKey==pOp->nOpt && !sqlite3_strnicmp(z, pOp->zOpt, pOp->nOpt) ){
1185 break;
1186 }
1187 }
1188 if( iOpt==SizeofArray(aFts4Opt) ){
1189 *pzErr = sqlite3_mprintf("unrecognized parameter: %s", z);
1190 rc = SQLITE_ERROR;
1191 }else{
1192 switch( iOpt ){
1193 case 0: /* MATCHINFO */
1194 if( strlen(zVal)!=4 || sqlite3_strnicmp(zVal, "fts3", 4) ){
1195 *pzErr = sqlite3_mprintf("unrecognized matchinfo: %s", zVal);
1196 rc = SQLITE_ERROR;
1197 }
1198 bNoDocsize = 1;
1199 break;
1200
1201 case 1: /* PREFIX */
1202 sqlite3_free(zPrefix);
1203 zPrefix = zVal;
1204 zVal = 0;
1205 break;
1206
1207 case 2: /* COMPRESS */
1208 sqlite3_free(zCompress);
1209 zCompress = zVal;
1210 zVal = 0;
1211 break;
1212
1213 case 3: /* UNCOMPRESS */
1214 sqlite3_free(zUncompress);
1215 zUncompress = zVal;
1216 zVal = 0;
1217 break;
1218
1219 case 4: /* ORDER */
1220 if( (strlen(zVal)!=3 || sqlite3_strnicmp(zVal, "asc", 3))
1221 && (strlen(zVal)!=4 || sqlite3_strnicmp(zVal, "desc", 4))
1222 ){
1223 *pzErr = sqlite3_mprintf("unrecognized order: %s", zVal);
1224 rc = SQLITE_ERROR;
1225 }
1226 bDescIdx = (zVal[0]=='d' || zVal[0]=='D');
1227 break;
1228
1229 case 5: /* CONTENT */
1230 sqlite3_free(zContent);
1231 zContent = zVal;
1232 zVal = 0;
1233 break;
1234
1235 case 6: /* LANGUAGEID */
1236 assert( iOpt==6 );
1237 sqlite3_free(zLanguageid);
1238 zLanguageid = zVal;
1239 zVal = 0;
1240 break;
1241
1242 case 7: /* NOTINDEXED */
1243 azNotindexed[nNotindexed++] = zVal;
1244 zVal = 0;
1245 break;
1246 }
1247 }
1248 sqlite3_free(zVal);
890 } 1249 }
891 if( nKey==9 && 0==sqlite3_strnicmp(z, "matchinfo", 9) ){
892 if( strlen(zVal)==4 && 0==sqlite3_strnicmp(zVal, "fts3", 4) ){
893 bNoDocsize = 1;
894 }else{
895 *pzErr = sqlite3_mprintf("unrecognized matchinfo: %s", zVal);
896 rc = SQLITE_ERROR;
897 }
898 }else if( nKey==8 && 0==sqlite3_strnicmp(z, "compress", 8) ){
899 zCompress = zVal;
900 zVal = 0;
901 }else if( nKey==10 && 0==sqlite3_strnicmp(z, "uncompress", 10) ){
902 zUncompress = zVal;
903 zVal = 0;
904 }else{
905 *pzErr = sqlite3_mprintf("unrecognized parameter: %s", z);
906 rc = SQLITE_ERROR;
907 }
908 sqlite3_free(zVal);
909 } 1250 }
910 1251
911 /* Otherwise, the argument is a column name. */ 1252 /* Otherwise, the argument is a column name. */
912 else { 1253 else {
913 nString += (int)(strlen(z) + 1); 1254 nString += (int)(strlen(z) + 1);
914 aCol[nCol++] = z; 1255 aCol[nCol++] = z;
915 } 1256 }
916 } 1257 }
1258
1259 /* If a content=xxx option was specified, the following:
1260 **
1261 ** 1. Ignore any compress= and uncompress= options.
1262 **
1263 ** 2. If no column names were specified as part of the CREATE VIRTUAL
1264 ** TABLE statement, use all columns from the content table.
1265 */
1266 if( rc==SQLITE_OK && zContent ){
1267 sqlite3_free(zCompress);
1268 sqlite3_free(zUncompress);
1269 zCompress = 0;
1270 zUncompress = 0;
1271 if( nCol==0 ){
1272 sqlite3_free((void*)aCol);
1273 aCol = 0;
1274 rc = fts3ContentColumns(db, argv[1], zContent, &aCol, &nCol, &nString);
1275
1276 /* If a languageid= option was specified, remove the language id
1277 ** column from the aCol[] array. */
1278 if( rc==SQLITE_OK && zLanguageid ){
1279 int j;
1280 for(j=0; j<nCol; j++){
1281 if( sqlite3_stricmp(zLanguageid, aCol[j])==0 ){
1282 int k;
1283 for(k=j; k<nCol; k++) aCol[k] = aCol[k+1];
1284 nCol--;
1285 break;
1286 }
1287 }
1288 }
1289 }
1290 }
917 if( rc!=SQLITE_OK ) goto fts3_init_out; 1291 if( rc!=SQLITE_OK ) goto fts3_init_out;
918 1292
919 if( nCol==0 ){ 1293 if( nCol==0 ){
920 assert( nString==0 ); 1294 assert( nString==0 );
921 aCol[0] = "content"; 1295 aCol[0] = "content";
922 nString = 8; 1296 nString = 8;
923 nCol = 1; 1297 nCol = 1;
924 } 1298 }
925 1299
926 if( pTokenizer==0 ){ 1300 if( pTokenizer==0 ){
927 rc = sqlite3Fts3InitTokenizer(pHash, "simple", &pTokenizer, pzErr); 1301 rc = sqlite3Fts3InitTokenizer(pHash, "simple", &pTokenizer, pzErr);
928 if( rc!=SQLITE_OK ) goto fts3_init_out; 1302 if( rc!=SQLITE_OK ) goto fts3_init_out;
929 } 1303 }
930 assert( pTokenizer ); 1304 assert( pTokenizer );
931 1305
1306 rc = fts3PrefixParameter(zPrefix, &nIndex, &aIndex);
1307 if( rc==SQLITE_ERROR ){
1308 assert( zPrefix );
1309 *pzErr = sqlite3_mprintf("error parsing prefix parameter: %s", zPrefix);
1310 }
1311 if( rc!=SQLITE_OK ) goto fts3_init_out;
932 1312
933 /* Allocate and populate the Fts3Table structure. */ 1313 /* Allocate and populate the Fts3Table structure. */
934 nByte = sizeof(Fts3Table) + /* Fts3Table */ 1314 nByte = sizeof(Fts3Table) + /* Fts3Table */
935 nCol * sizeof(char *) + /* azColumn */ 1315 nCol * sizeof(char *) + /* azColumn */
1316 nIndex * sizeof(struct Fts3Index) + /* aIndex */
1317 nCol * sizeof(u8) + /* abNotindexed */
936 nName + /* zName */ 1318 nName + /* zName */
937 nDb + /* zDb */ 1319 nDb + /* zDb */
938 nString; /* Space for azColumn strings */ 1320 nString; /* Space for azColumn strings */
939 p = (Fts3Table*)sqlite3_malloc(nByte); 1321 p = (Fts3Table*)sqlite3_malloc(nByte);
940 if( p==0 ){ 1322 if( p==0 ){
941 rc = SQLITE_NOMEM; 1323 rc = SQLITE_NOMEM;
942 goto fts3_init_out; 1324 goto fts3_init_out;
943 } 1325 }
944 memset(p, 0, nByte); 1326 memset(p, 0, nByte);
945 p->db = db; 1327 p->db = db;
946 p->nColumn = nCol; 1328 p->nColumn = nCol;
947 p->nPendingData = 0; 1329 p->nPendingData = 0;
948 p->azColumn = (char **)&p[1]; 1330 p->azColumn = (char **)&p[1];
949 p->pTokenizer = pTokenizer; 1331 p->pTokenizer = pTokenizer;
950 p->nNodeSize = 1000;
951 p->nMaxPendingData = FTS3_MAX_PENDING_DATA; 1332 p->nMaxPendingData = FTS3_MAX_PENDING_DATA;
952 p->bHasDocsize = (isFts4 && bNoDocsize==0); 1333 p->bHasDocsize = (isFts4 && bNoDocsize==0);
953 p->bHasStat = isFts4; 1334 p->bHasStat = isFts4;
954 fts3HashInit(&p->pendingTerms, FTS3_HASH_STRING, 1); 1335 p->bFts4 = isFts4;
1336 p->bDescIdx = bDescIdx;
1337 p->nAutoincrmerge = 0xff; /* 0xff means setting unknown */
1338 p->zContentTbl = zContent;
1339 p->zLanguageid = zLanguageid;
1340 zContent = 0;
1341 zLanguageid = 0;
1342 TESTONLY( p->inTransaction = -1 );
1343 TESTONLY( p->mxSavepoint = -1 );
1344
1345 p->aIndex = (struct Fts3Index *)&p->azColumn[nCol];
1346 memcpy(p->aIndex, aIndex, sizeof(struct Fts3Index) * nIndex);
1347 p->nIndex = nIndex;
1348 for(i=0; i<nIndex; i++){
1349 fts3HashInit(&p->aIndex[i].hPending, FTS3_HASH_STRING, 1);
1350 }
1351 p->abNotindexed = (u8 *)&p->aIndex[nIndex];
955 1352
956 /* Fill in the zName and zDb fields of the vtab structure. */ 1353 /* Fill in the zName and zDb fields of the vtab structure. */
957 zCsr = (char *)&p->azColumn[nCol]; 1354 zCsr = (char *)&p->abNotindexed[nCol];
958 p->zName = zCsr; 1355 p->zName = zCsr;
959 memcpy(zCsr, argv[2], nName); 1356 memcpy(zCsr, argv[2], nName);
960 zCsr += nName; 1357 zCsr += nName;
961 p->zDb = zCsr; 1358 p->zDb = zCsr;
962 memcpy(zCsr, argv[1], nDb); 1359 memcpy(zCsr, argv[1], nDb);
963 zCsr += nDb; 1360 zCsr += nDb;
964 1361
965 /* Fill in the azColumn array */ 1362 /* Fill in the azColumn array */
966 for(iCol=0; iCol<nCol; iCol++){ 1363 for(iCol=0; iCol<nCol; iCol++){
967 char *z; 1364 char *z;
968 int n; 1365 int n = 0;
969 z = (char *)sqlite3Fts3NextToken(aCol[iCol], &n); 1366 z = (char *)sqlite3Fts3NextToken(aCol[iCol], &n);
970 memcpy(zCsr, z, n); 1367 memcpy(zCsr, z, n);
971 zCsr[n] = '\0'; 1368 zCsr[n] = '\0';
972 sqlite3Fts3Dequote(zCsr); 1369 sqlite3Fts3Dequote(zCsr);
973 p->azColumn[iCol] = zCsr; 1370 p->azColumn[iCol] = zCsr;
974 zCsr += n+1; 1371 zCsr += n+1;
975 assert( zCsr <= &((char *)p)[nByte] ); 1372 assert( zCsr <= &((char *)p)[nByte] );
976 } 1373 }
977 1374
978 if( (zCompress==0)!=(zUncompress==0) ){ 1375 /* Fill in the abNotindexed array */
1376 for(iCol=0; iCol<nCol; iCol++){
1377 int n = (int)strlen(p->azColumn[iCol]);
1378 for(i=0; i<nNotindexed; i++){
1379 char *zNot = azNotindexed[i];
1380 if( zNot && n==(int)strlen(zNot)
1381 && 0==sqlite3_strnicmp(p->azColumn[iCol], zNot, n)
1382 ){
1383 p->abNotindexed[iCol] = 1;
1384 sqlite3_free(zNot);
1385 azNotindexed[i] = 0;
1386 }
1387 }
1388 }
1389 for(i=0; i<nNotindexed; i++){
1390 if( azNotindexed[i] ){
1391 *pzErr = sqlite3_mprintf("no such column: %s", azNotindexed[i]);
1392 rc = SQLITE_ERROR;
1393 }
1394 }
1395
1396 if( rc==SQLITE_OK && (zCompress==0)!=(zUncompress==0) ){
979 char const *zMiss = (zCompress==0 ? "compress" : "uncompress"); 1397 char const *zMiss = (zCompress==0 ? "compress" : "uncompress");
980 rc = SQLITE_ERROR; 1398 rc = SQLITE_ERROR;
981 *pzErr = sqlite3_mprintf("missing %s parameter in fts4 constructor", zMiss); 1399 *pzErr = sqlite3_mprintf("missing %s parameter in fts4 constructor", zMiss);
982 } 1400 }
983 p->zReadExprlist = fts3ReadExprList(p, zUncompress, &rc); 1401 p->zReadExprlist = fts3ReadExprList(p, zUncompress, &rc);
984 p->zWriteExprlist = fts3WriteExprList(p, zCompress, &rc); 1402 p->zWriteExprlist = fts3WriteExprList(p, zCompress, &rc);
985 if( rc!=SQLITE_OK ) goto fts3_init_out; 1403 if( rc!=SQLITE_OK ) goto fts3_init_out;
986 1404
987 /* If this is an xCreate call, create the underlying tables in the 1405 /* If this is an xCreate call, create the underlying tables in the
988 ** database. TODO: For xConnect(), it could verify that said tables exist. 1406 ** database. TODO: For xConnect(), it could verify that said tables exist.
989 */ 1407 */
990 if( isCreate ){ 1408 if( isCreate ){
991 rc = fts3CreateTables(p); 1409 rc = fts3CreateTables(p);
992 } 1410 }
993 1411
1412 /* Check to see if a legacy fts3 table has been "upgraded" by the
1413 ** addition of a %_stat table so that it can use incremental merge.
1414 */
1415 if( !isFts4 && !isCreate ){
1416 p->bHasStat = 2;
1417 }
1418
994 /* Figure out the page-size for the database. This is required in order to 1419 /* Figure out the page-size for the database. This is required in order to
995 ** estimate the cost of loading large doclists from the database (see 1420 ** estimate the cost of loading large doclists from the database. */
996 ** function sqlite3Fts3SegReaderCost() for details).
997 */
998 fts3DatabasePageSize(&rc, p); 1421 fts3DatabasePageSize(&rc, p);
1422 p->nNodeSize = p->nPgsz-35;
999 1423
1000 /* Declare the table schema to SQLite. */ 1424 /* Declare the table schema to SQLite. */
1001 fts3DeclareVtab(&rc, p); 1425 fts3DeclareVtab(&rc, p);
1002 1426
1003 fts3_init_out: 1427 fts3_init_out:
1428 sqlite3_free(zPrefix);
1429 sqlite3_free(aIndex);
1004 sqlite3_free(zCompress); 1430 sqlite3_free(zCompress);
1005 sqlite3_free(zUncompress); 1431 sqlite3_free(zUncompress);
1432 sqlite3_free(zContent);
1433 sqlite3_free(zLanguageid);
1434 for(i=0; i<nNotindexed; i++) sqlite3_free(azNotindexed[i]);
1006 sqlite3_free((void *)aCol); 1435 sqlite3_free((void *)aCol);
1436 sqlite3_free((void *)azNotindexed);
1007 if( rc!=SQLITE_OK ){ 1437 if( rc!=SQLITE_OK ){
1008 if( p ){ 1438 if( p ){
1009 fts3DisconnectMethod((sqlite3_vtab *)p); 1439 fts3DisconnectMethod((sqlite3_vtab *)p);
1010 }else if( pTokenizer ){ 1440 }else if( pTokenizer ){
1011 pTokenizer->pModule->xDestroy(pTokenizer); 1441 pTokenizer->pModule->xDestroy(pTokenizer);
1012 } 1442 }
1013 }else{ 1443 }else{
1444 assert( p->pSegments==0 );
1014 *ppVTab = &p->base; 1445 *ppVTab = &p->base;
1015 } 1446 }
1016 return rc; 1447 return rc;
1017 } 1448 }
1018 1449
1019 /* 1450 /*
1020 ** The xConnect() and xCreate() methods for the virtual table. All the 1451 ** The xConnect() and xCreate() methods for the virtual table. All the
1021 ** work is done in function fts3InitVtab(). 1452 ** work is done in function fts3InitVtab().
1022 */ 1453 */
1023 static int fts3ConnectMethod( 1454 static int fts3ConnectMethod(
(...skipping 10 matching lines...) Expand all
1034 sqlite3 *db, /* Database connection */ 1465 sqlite3 *db, /* Database connection */
1035 void *pAux, /* Pointer to tokenizer hash table */ 1466 void *pAux, /* Pointer to tokenizer hash table */
1036 int argc, /* Number of elements in argv array */ 1467 int argc, /* Number of elements in argv array */
1037 const char * const *argv, /* xCreate/xConnect argument array */ 1468 const char * const *argv, /* xCreate/xConnect argument array */
1038 sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */ 1469 sqlite3_vtab **ppVtab, /* OUT: New sqlite3_vtab object */
1039 char **pzErr /* OUT: sqlite3_malloc'd error message */ 1470 char **pzErr /* OUT: sqlite3_malloc'd error message */
1040 ){ 1471 ){
1041 return fts3InitVtab(1, db, pAux, argc, argv, ppVtab, pzErr); 1472 return fts3InitVtab(1, db, pAux, argc, argv, ppVtab, pzErr);
1042 } 1473 }
1043 1474
1475 /*
1476 ** Set the pIdxInfo->estimatedRows variable to nRow. Unless this
1477 ** extension is currently being used by a version of SQLite too old to
1478 ** support estimatedRows. In that case this function is a no-op.
1479 */
1480 static void fts3SetEstimatedRows(sqlite3_index_info *pIdxInfo, i64 nRow){
1481 #if SQLITE_VERSION_NUMBER>=3008002
1482 if( sqlite3_libversion_number()>=3008002 ){
1483 pIdxInfo->estimatedRows = nRow;
1484 }
1485 #endif
1486 }
1487
1044 /* 1488 /*
1045 ** Implementation of the xBestIndex method for FTS3 tables. There 1489 ** Implementation of the xBestIndex method for FTS3 tables. There
1046 ** are three possible strategies, in order of preference: 1490 ** are three possible strategies, in order of preference:
1047 ** 1491 **
1048 ** 1. Direct lookup by rowid or docid. 1492 ** 1. Direct lookup by rowid or docid.
1049 ** 2. Full-text search using a MATCH operator on a non-docid column. 1493 ** 2. Full-text search using a MATCH operator on a non-docid column.
1050 ** 3. Linear scan of %_content table. 1494 ** 3. Linear scan of %_content table.
1051 */ 1495 */
1052 static int fts3BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){ 1496 static int fts3BestIndexMethod(sqlite3_vtab *pVTab, sqlite3_index_info *pInfo){
1053 Fts3Table *p = (Fts3Table *)pVTab; 1497 Fts3Table *p = (Fts3Table *)pVTab;
1054 int i; /* Iterator variable */ 1498 int i; /* Iterator variable */
1055 int iCons = -1; /* Index of constraint to use */ 1499 int iCons = -1; /* Index of constraint to use */
1056 1500
1501 int iLangidCons = -1; /* Index of langid=x constraint, if present */
1502 int iDocidGe = -1; /* Index of docid>=x constraint, if present */
1503 int iDocidLe = -1; /* Index of docid<=x constraint, if present */
1504 int iIdx;
1505
1057 /* By default use a full table scan. This is an expensive option, 1506 /* By default use a full table scan. This is an expensive option,
1058 ** so search through the constraints to see if a more efficient 1507 ** so search through the constraints to see if a more efficient
1059 ** strategy is possible. 1508 ** strategy is possible.
1060 */ 1509 */
1061 pInfo->idxNum = FTS3_FULLSCAN_SEARCH; 1510 pInfo->idxNum = FTS3_FULLSCAN_SEARCH;
1062 pInfo->estimatedCost = 500000; 1511 pInfo->estimatedCost = 5000000;
1063 for(i=0; i<pInfo->nConstraint; i++){ 1512 for(i=0; i<pInfo->nConstraint; i++){
1513 int bDocid; /* True if this constraint is on docid */
1064 struct sqlite3_index_constraint *pCons = &pInfo->aConstraint[i]; 1514 struct sqlite3_index_constraint *pCons = &pInfo->aConstraint[i];
1065 if( pCons->usable==0 ) continue; 1515 if( pCons->usable==0 ){
1516 if( pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH ){
1517 /* There exists an unusable MATCH constraint. This means that if
1518 ** the planner does elect to use the results of this call as part
1519 ** of the overall query plan the user will see an "unable to use
1520 ** function MATCH in the requested context" error. To discourage
1521 ** this, return a very high cost here. */
1522 pInfo->idxNum = FTS3_FULLSCAN_SEARCH;
1523 pInfo->estimatedCost = 1e50;
1524 fts3SetEstimatedRows(pInfo, ((sqlite3_int64)1) << 50);
1525 return SQLITE_OK;
1526 }
1527 continue;
1528 }
1529
1530 bDocid = (pCons->iColumn<0 || pCons->iColumn==p->nColumn+1);
1066 1531
1067 /* A direct lookup on the rowid or docid column. Assign a cost of 1.0. */ 1532 /* A direct lookup on the rowid or docid column. Assign a cost of 1.0. */
1068 if( pCons->op==SQLITE_INDEX_CONSTRAINT_EQ 1533 if( iCons<0 && pCons->op==SQLITE_INDEX_CONSTRAINT_EQ && bDocid ){
1069 && (pCons->iColumn<0 || pCons->iColumn==p->nColumn+1 )
1070 ){
1071 pInfo->idxNum = FTS3_DOCID_SEARCH; 1534 pInfo->idxNum = FTS3_DOCID_SEARCH;
1072 pInfo->estimatedCost = 1.0; 1535 pInfo->estimatedCost = 1.0;
1073 iCons = i; 1536 iCons = i;
1074 } 1537 }
1075 1538
1076 /* A MATCH constraint. Use a full-text search. 1539 /* A MATCH constraint. Use a full-text search.
1077 ** 1540 **
1078 ** If there is more than one MATCH constraint available, use the first 1541 ** If there is more than one MATCH constraint available, use the first
1079 ** one encountered. If there is both a MATCH constraint and a direct 1542 ** one encountered. If there is both a MATCH constraint and a direct
1080 ** rowid/docid lookup, prefer the MATCH strategy. This is done even 1543 ** rowid/docid lookup, prefer the MATCH strategy. This is done even
1081 ** though the rowid/docid lookup is faster than a MATCH query, selecting 1544 ** though the rowid/docid lookup is faster than a MATCH query, selecting
1082 ** it would lead to an "unable to use function MATCH in the requested 1545 ** it would lead to an "unable to use function MATCH in the requested
1083 ** context" error. 1546 ** context" error.
1084 */ 1547 */
1085 if( pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH 1548 if( pCons->op==SQLITE_INDEX_CONSTRAINT_MATCH
1086 && pCons->iColumn>=0 && pCons->iColumn<=p->nColumn 1549 && pCons->iColumn>=0 && pCons->iColumn<=p->nColumn
1087 ){ 1550 ){
1088 pInfo->idxNum = FTS3_FULLTEXT_SEARCH + pCons->iColumn; 1551 pInfo->idxNum = FTS3_FULLTEXT_SEARCH + pCons->iColumn;
1089 pInfo->estimatedCost = 2.0; 1552 pInfo->estimatedCost = 2.0;
1090 iCons = i; 1553 iCons = i;
1091 break; 1554 }
1555
1556 /* Equality constraint on the langid column */
1557 if( pCons->op==SQLITE_INDEX_CONSTRAINT_EQ
1558 && pCons->iColumn==p->nColumn + 2
1559 ){
1560 iLangidCons = i;
1561 }
1562
1563 if( bDocid ){
1564 switch( pCons->op ){
1565 case SQLITE_INDEX_CONSTRAINT_GE:
1566 case SQLITE_INDEX_CONSTRAINT_GT:
1567 iDocidGe = i;
1568 break;
1569
1570 case SQLITE_INDEX_CONSTRAINT_LE:
1571 case SQLITE_INDEX_CONSTRAINT_LT:
1572 iDocidLe = i;
1573 break;
1574 }
1092 } 1575 }
1093 } 1576 }
1094 1577
1578 iIdx = 1;
1095 if( iCons>=0 ){ 1579 if( iCons>=0 ){
1096 pInfo->aConstraintUsage[iCons].argvIndex = 1; 1580 pInfo->aConstraintUsage[iCons].argvIndex = iIdx++;
1097 pInfo->aConstraintUsage[iCons].omit = 1; 1581 pInfo->aConstraintUsage[iCons].omit = 1;
1098 } 1582 }
1583 if( iLangidCons>=0 ){
1584 pInfo->idxNum |= FTS3_HAVE_LANGID;
1585 pInfo->aConstraintUsage[iLangidCons].argvIndex = iIdx++;
1586 }
1587 if( iDocidGe>=0 ){
1588 pInfo->idxNum |= FTS3_HAVE_DOCID_GE;
1589 pInfo->aConstraintUsage[iDocidGe].argvIndex = iIdx++;
1590 }
1591 if( iDocidLe>=0 ){
1592 pInfo->idxNum |= FTS3_HAVE_DOCID_LE;
1593 pInfo->aConstraintUsage[iDocidLe].argvIndex = iIdx++;
1594 }
1595
1596 /* Regardless of the strategy selected, FTS can deliver rows in rowid (or
1597 ** docid) order. Both ascending and descending are possible.
1598 */
1599 if( pInfo->nOrderBy==1 ){
1600 struct sqlite3_index_orderby *pOrder = &pInfo->aOrderBy[0];
1601 if( pOrder->iColumn<0 || pOrder->iColumn==p->nColumn+1 ){
1602 if( pOrder->desc ){
1603 pInfo->idxStr = "DESC";
1604 }else{
1605 pInfo->idxStr = "ASC";
1606 }
1607 pInfo->orderByConsumed = 1;
1608 }
1609 }
1610
1611 assert( p->pSegments==0 );
1099 return SQLITE_OK; 1612 return SQLITE_OK;
1100 } 1613 }
1101 1614
1102 /* 1615 /*
1103 ** Implementation of xOpen method. 1616 ** Implementation of xOpen method.
1104 */ 1617 */
1105 static int fts3OpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){ 1618 static int fts3OpenMethod(sqlite3_vtab *pVTab, sqlite3_vtab_cursor **ppCsr){
1106 sqlite3_vtab_cursor *pCsr; /* Allocated cursor */ 1619 sqlite3_vtab_cursor *pCsr; /* Allocated cursor */
1107 1620
1108 UNUSED_PARAMETER(pVTab); 1621 UNUSED_PARAMETER(pVTab);
(...skipping 15 matching lines...) Expand all
1124 ** on the xClose method of the virtual table interface. 1637 ** on the xClose method of the virtual table interface.
1125 */ 1638 */
1126 static int fts3CloseMethod(sqlite3_vtab_cursor *pCursor){ 1639 static int fts3CloseMethod(sqlite3_vtab_cursor *pCursor){
1127 Fts3Cursor *pCsr = (Fts3Cursor *)pCursor; 1640 Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;
1128 assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 ); 1641 assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
1129 sqlite3_finalize(pCsr->pStmt); 1642 sqlite3_finalize(pCsr->pStmt);
1130 sqlite3Fts3ExprFree(pCsr->pExpr); 1643 sqlite3Fts3ExprFree(pCsr->pExpr);
1131 sqlite3Fts3FreeDeferredTokens(pCsr); 1644 sqlite3Fts3FreeDeferredTokens(pCsr);
1132 sqlite3_free(pCsr->aDoclist); 1645 sqlite3_free(pCsr->aDoclist);
1133 sqlite3_free(pCsr->aMatchinfo); 1646 sqlite3_free(pCsr->aMatchinfo);
1647 assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
1134 sqlite3_free(pCsr); 1648 sqlite3_free(pCsr);
1135 return SQLITE_OK; 1649 return SQLITE_OK;
1136 } 1650 }
1137 1651
1138 /* 1652 /*
1653 ** If pCsr->pStmt has not been prepared (i.e. if pCsr->pStmt==0), then
1654 ** compose and prepare an SQL statement of the form:
1655 **
1656 ** "SELECT <columns> FROM %_content WHERE rowid = ?"
1657 **
1658 ** (or the equivalent for a content=xxx table) and set pCsr->pStmt to
1659 ** it. If an error occurs, return an SQLite error code.
1660 **
1661 ** Otherwise, set *ppStmt to point to pCsr->pStmt and return SQLITE_OK.
1662 */
1663 static int fts3CursorSeekStmt(Fts3Cursor *pCsr, sqlite3_stmt **ppStmt){
1664 int rc = SQLITE_OK;
1665 if( pCsr->pStmt==0 ){
1666 Fts3Table *p = (Fts3Table *)pCsr->base.pVtab;
1667 char *zSql;
1668 zSql = sqlite3_mprintf("SELECT %s WHERE rowid = ?", p->zReadExprlist);
1669 if( !zSql ) return SQLITE_NOMEM;
1670 rc = sqlite3_prepare_v2(p->db, zSql, -1, &pCsr->pStmt, 0);
1671 sqlite3_free(zSql);
1672 }
1673 *ppStmt = pCsr->pStmt;
1674 return rc;
1675 }
1676
1677 /*
1139 ** Position the pCsr->pStmt statement so that it is on the row 1678 ** Position the pCsr->pStmt statement so that it is on the row
1140 ** of the %_content table that contains the last match. Return 1679 ** of the %_content table that contains the last match. Return
1141 ** SQLITE_OK on success. 1680 ** SQLITE_OK on success.
1142 */ 1681 */
1143 static int fts3CursorSeek(sqlite3_context *pContext, Fts3Cursor *pCsr){ 1682 static int fts3CursorSeek(sqlite3_context *pContext, Fts3Cursor *pCsr){
1683 int rc = SQLITE_OK;
1144 if( pCsr->isRequireSeek ){ 1684 if( pCsr->isRequireSeek ){
1145 pCsr->isRequireSeek = 0; 1685 sqlite3_stmt *pStmt = 0;
1146 sqlite3_bind_int64(pCsr->pStmt, 1, pCsr->iPrevId); 1686
1147 if( SQLITE_ROW==sqlite3_step(pCsr->pStmt) ){ 1687 rc = fts3CursorSeekStmt(pCsr, &pStmt);
1148 return SQLITE_OK; 1688 if( rc==SQLITE_OK ){
1149 }else{ 1689 sqlite3_bind_int64(pCsr->pStmt, 1, pCsr->iPrevId);
1150 int rc = sqlite3_reset(pCsr->pStmt); 1690 pCsr->isRequireSeek = 0;
1151 if( rc==SQLITE_OK ){ 1691 if( SQLITE_ROW==sqlite3_step(pCsr->pStmt) ){
1152 /* If no row was found and no error has occured, then the %_content 1692 return SQLITE_OK;
1153 ** table is missing a row that is present in the full-text index. 1693 }else{
1154 ** The data structures are corrupt. 1694 rc = sqlite3_reset(pCsr->pStmt);
1155 */ 1695 if( rc==SQLITE_OK && ((Fts3Table *)pCsr->base.pVtab)->zContentTbl==0 ){
1156 rc = SQLITE_CORRUPT; 1696 /* If no row was found and no error has occurred, then the %_content
1697 ** table is missing a row that is present in the full-text index.
1698 ** The data structures are corrupt. */
1699 rc = FTS_CORRUPT_VTAB;
1700 pCsr->isEof = 1;
1701 }
1157 } 1702 }
1158 pCsr->isEof = 1;
1159 if( pContext ){
1160 sqlite3_result_error_code(pContext, rc);
1161 }
1162 return rc;
1163 } 1703 }
1164 }else{
1165 return SQLITE_OK;
1166 } 1704 }
1705
1706 if( rc!=SQLITE_OK && pContext ){
1707 sqlite3_result_error_code(pContext, rc);
1708 }
1709 return rc;
1167 } 1710 }
1168 1711
1169 /* 1712 /*
1170 ** This function is used to process a single interior node when searching 1713 ** This function is used to process a single interior node when searching
1171 ** a b-tree for a term or term prefix. The node data is passed to this 1714 ** a b-tree for a term or term prefix. The node data is passed to this
1172 ** function via the zNode/nNode parameters. The term to search for is 1715 ** function via the zNode/nNode parameters. The term to search for is
1173 ** passed in zTerm/nTerm. 1716 ** passed in zTerm/nTerm.
1174 ** 1717 **
1175 ** If piFirst is not NULL, then this function sets *piFirst to the blockid 1718 ** If piFirst is not NULL, then this function sets *piFirst to the blockid
1176 ** of the child node that heads the sub-tree that may contain the term. 1719 ** of the child node that heads the sub-tree that may contain the term.
(...skipping 29 matching lines...) Expand all
1206 ** root node, then the buffer comes from a SELECT statement. SQLite does 1749 ** root node, then the buffer comes from a SELECT statement. SQLite does
1207 ** not make this guarantee explicitly, but in practice there are always 1750 ** not make this guarantee explicitly, but in practice there are always
1208 ** either more than 20 bytes of allocated space following the nNode bytes of 1751 ** either more than 20 bytes of allocated space following the nNode bytes of
1209 ** contents, or two zero bytes. Or, if the node is read from the %_segments 1752 ** contents, or two zero bytes. Or, if the node is read from the %_segments
1210 ** table, then there are always 20 bytes of zeroed padding following the 1753 ** table, then there are always 20 bytes of zeroed padding following the
1211 ** nNode bytes of content (see sqlite3Fts3ReadBlock() for details). 1754 ** nNode bytes of content (see sqlite3Fts3ReadBlock() for details).
1212 */ 1755 */
1213 zCsr += sqlite3Fts3GetVarint(zCsr, &iChild); 1756 zCsr += sqlite3Fts3GetVarint(zCsr, &iChild);
1214 zCsr += sqlite3Fts3GetVarint(zCsr, &iChild); 1757 zCsr += sqlite3Fts3GetVarint(zCsr, &iChild);
1215 if( zCsr>zEnd ){ 1758 if( zCsr>zEnd ){
1216 return SQLITE_CORRUPT; 1759 return FTS_CORRUPT_VTAB;
1217 } 1760 }
1218 1761
1219 while( zCsr<zEnd && (piFirst || piLast) ){ 1762 while( zCsr<zEnd && (piFirst || piLast) ){
1220 int cmp; /* memcmp() result */ 1763 int cmp; /* memcmp() result */
1221 int nSuffix; /* Size of term suffix */ 1764 int nSuffix; /* Size of term suffix */
1222 int nPrefix = 0; /* Size of term prefix */ 1765 int nPrefix = 0; /* Size of term prefix */
1223 int nBuffer; /* Total term size */ 1766 int nBuffer; /* Total term size */
1224 1767
1225 /* Load the next term on the node into zBuffer. Use realloc() to expand 1768 /* Load the next term on the node into zBuffer. Use realloc() to expand
1226 ** the size of zBuffer if required. */ 1769 ** the size of zBuffer if required. */
1227 if( !isFirstTerm ){ 1770 if( !isFirstTerm ){
1228 zCsr += sqlite3Fts3GetVarint32(zCsr, &nPrefix); 1771 zCsr += fts3GetVarint32(zCsr, &nPrefix);
1229 } 1772 }
1230 isFirstTerm = 0; 1773 isFirstTerm = 0;
1231 zCsr += sqlite3Fts3GetVarint32(zCsr, &nSuffix); 1774 zCsr += fts3GetVarint32(zCsr, &nSuffix);
1232 1775
1233 /* NOTE(shess): Previous code checked for negative nPrefix and 1776 /* NOTE(shess): Previous code checked for negative nPrefix and
1234 ** nSuffix and suffix overrunning zEnd. Additionally corrupt if 1777 ** nSuffix and suffix overrunning zEnd. Additionally corrupt if
1235 ** the prefix is longer than the previous term, or if the suffix 1778 ** the prefix is longer than the previous term, or if the suffix
1236 ** causes overflow. 1779 ** causes overflow.
1237 */ 1780 */
1238 if( nPrefix<0 || nSuffix<0 /* || nPrefix>nBuffer */ 1781 if( nPrefix<0 || nSuffix<0 /* || nPrefix>nBuffer */
1239 || &zCsr[nSuffix]<zCsr || &zCsr[nSuffix]>zEnd ){ 1782 || &zCsr[nSuffix]<zCsr || &zCsr[nSuffix]>zEnd ){
1240 rc = SQLITE_CORRUPT; 1783 rc = SQLITE_CORRUPT;
1241 goto finish_scan; 1784 goto finish_scan;
1242 } 1785 }
1243 if( nPrefix+nSuffix>nAlloc ){ 1786 if( nPrefix+nSuffix>nAlloc ){
1244 char *zNew; 1787 char *zNew;
1245 nAlloc = (nPrefix+nSuffix) * 2; 1788 nAlloc = (nPrefix+nSuffix) * 2;
1246 zNew = (char *)sqlite3_realloc(zBuffer, nAlloc); 1789 zNew = (char *)sqlite3_realloc(zBuffer, nAlloc);
1247 if( !zNew ){ 1790 if( !zNew ){
1248 rc = SQLITE_NOMEM; 1791 rc = SQLITE_NOMEM;
1249 goto finish_scan; 1792 goto finish_scan;
1250 } 1793 }
1251 zBuffer = zNew; 1794 zBuffer = zNew;
1252 } 1795 }
1796 assert( zBuffer );
1253 memcpy(&zBuffer[nPrefix], zCsr, nSuffix); 1797 memcpy(&zBuffer[nPrefix], zCsr, nSuffix);
1254 nBuffer = nPrefix + nSuffix; 1798 nBuffer = nPrefix + nSuffix;
1255 zCsr += nSuffix; 1799 zCsr += nSuffix;
1256 1800
1257 /* Compare the term we are searching for with the term just loaded from 1801 /* Compare the term we are searching for with the term just loaded from
1258 ** the interior node. If the specified term is greater than or equal 1802 ** the interior node. If the specified term is greater than or equal
1259 ** to the term from the interior node, then all terms on the sub-tree 1803 ** to the term from the interior node, then all terms on the sub-tree
1260 ** headed by node iChild are smaller than zTerm. No need to search 1804 ** headed by node iChild are smaller than zTerm. No need to search
1261 ** iChild. 1805 ** iChild.
1262 ** 1806 **
(...skipping 51 matching lines...) Expand 10 before | Expand all | Expand 10 after
1314 const char *zNode, /* Buffer containing segment interior node */ 1858 const char *zNode, /* Buffer containing segment interior node */
1315 int nNode, /* Size of buffer at zNode */ 1859 int nNode, /* Size of buffer at zNode */
1316 sqlite3_int64 *piLeaf, /* Selected leaf node */ 1860 sqlite3_int64 *piLeaf, /* Selected leaf node */
1317 sqlite3_int64 *piLeaf2 /* Selected leaf node */ 1861 sqlite3_int64 *piLeaf2 /* Selected leaf node */
1318 ){ 1862 ){
1319 int rc; /* Return code */ 1863 int rc; /* Return code */
1320 int iHeight; /* Height of this node in tree */ 1864 int iHeight; /* Height of this node in tree */
1321 1865
1322 assert( piLeaf || piLeaf2 ); 1866 assert( piLeaf || piLeaf2 );
1323 1867
1324 sqlite3Fts3GetVarint32(zNode, &iHeight); 1868 fts3GetVarint32(zNode, &iHeight);
1325 rc = fts3ScanInteriorNode(zTerm, nTerm, zNode, nNode, piLeaf, piLeaf2); 1869 rc = fts3ScanInteriorNode(zTerm, nTerm, zNode, nNode, piLeaf, piLeaf2);
1326 assert( !piLeaf2 || !piLeaf || rc!=SQLITE_OK || (*piLeaf<=*piLeaf2) ); 1870 assert( !piLeaf2 || !piLeaf || rc!=SQLITE_OK || (*piLeaf<=*piLeaf2) );
1327 1871
1328 if( rc==SQLITE_OK && iHeight>1 ){ 1872 if( rc==SQLITE_OK && iHeight>1 ){
1329 char *zBlob = 0; /* Blob read from %_segments table */ 1873 char *zBlob = 0; /* Blob read from %_segments table */
1330 int nBlob; /* Size of zBlob in bytes */ 1874 int nBlob; /* Size of zBlob in bytes */
1331 1875
1332 if( piLeaf && piLeaf2 && (*piLeaf!=*piLeaf2) ){ 1876 if( piLeaf && piLeaf2 && (*piLeaf!=*piLeaf2) ){
1333 rc = sqlite3Fts3ReadBlock(p, *piLeaf, &zBlob, &nBlob); 1877 rc = sqlite3Fts3ReadBlock(p, *piLeaf, &zBlob, &nBlob, 0);
1334 if( rc==SQLITE_OK ){ 1878 if( rc==SQLITE_OK ){
1335 rc = fts3SelectLeaf(p, zTerm, nTerm, zBlob, nBlob, piLeaf, 0); 1879 rc = fts3SelectLeaf(p, zTerm, nTerm, zBlob, nBlob, piLeaf, 0);
1336 } 1880 }
1337 sqlite3_free(zBlob); 1881 sqlite3_free(zBlob);
1338 piLeaf = 0; 1882 piLeaf = 0;
1339 zBlob = 0; 1883 zBlob = 0;
1340 } 1884 }
1341 1885
1342 if( rc==SQLITE_OK ){ 1886 if( rc==SQLITE_OK ){
1343 rc = sqlite3Fts3ReadBlock(p, piLeaf ? *piLeaf : *piLeaf2, &zBlob, &nBlob); 1887 rc = sqlite3Fts3ReadBlock(p, piLeaf?*piLeaf:*piLeaf2, &zBlob, &nBlob, 0);
1344 } 1888 }
1345 if( rc==SQLITE_OK ){ 1889 if( rc==SQLITE_OK ){
1346 rc = fts3SelectLeaf(p, zTerm, nTerm, zBlob, nBlob, piLeaf, piLeaf2); 1890 rc = fts3SelectLeaf(p, zTerm, nTerm, zBlob, nBlob, piLeaf, piLeaf2);
1347 } 1891 }
1348 sqlite3_free(zBlob); 1892 sqlite3_free(zBlob);
1349 } 1893 }
1350 1894
1351 return rc; 1895 return rc;
1352 } 1896 }
1353 1897
(...skipping 162 matching lines...) Expand 10 before | Expand all | Expand 10 after
1516 char **pp2 /* Right input list */ 2060 char **pp2 /* Right input list */
1517 ){ 2061 ){
1518 char *p = *pp; 2062 char *p = *pp;
1519 char *p1 = *pp1; 2063 char *p1 = *pp1;
1520 char *p2 = *pp2; 2064 char *p2 = *pp2;
1521 2065
1522 while( *p1 || *p2 ){ 2066 while( *p1 || *p2 ){
1523 int iCol1; /* The current column index in pp1 */ 2067 int iCol1; /* The current column index in pp1 */
1524 int iCol2; /* The current column index in pp2 */ 2068 int iCol2; /* The current column index in pp2 */
1525 2069
1526 if( *p1==POS_COLUMN ) sqlite3Fts3GetVarint32(&p1[1], &iCol1); 2070 if( *p1==POS_COLUMN ) fts3GetVarint32(&p1[1], &iCol1);
1527 else if( *p1==POS_END ) iCol1 = POSITION_LIST_END; 2071 else if( *p1==POS_END ) iCol1 = POSITION_LIST_END;
1528 else iCol1 = 0; 2072 else iCol1 = 0;
1529 2073
1530 if( *p2==POS_COLUMN ) sqlite3Fts3GetVarint32(&p2[1], &iCol2); 2074 if( *p2==POS_COLUMN ) fts3GetVarint32(&p2[1], &iCol2);
1531 else if( *p2==POS_END ) iCol2 = POSITION_LIST_END; 2075 else if( *p2==POS_END ) iCol2 = POSITION_LIST_END;
1532 else iCol2 = 0; 2076 else iCol2 = 0;
1533 2077
1534 if( iCol1==iCol2 ){ 2078 if( iCol1==iCol2 ){
1535 sqlite3_int64 i1 = 0; /* Last position from pp1 */ 2079 sqlite3_int64 i1 = 0; /* Last position from pp1 */
1536 sqlite3_int64 i2 = 0; /* Last position from pp2 */ 2080 sqlite3_int64 i2 = 0; /* Last position from pp2 */
1537 sqlite3_int64 iPrev = 0; 2081 sqlite3_int64 iPrev = 0;
1538 int n = fts3PutColNumber(&p, iCol1); 2082 int n = fts3PutColNumber(&p, iCol1);
1539 p1 += n; 2083 p1 += n;
1540 p2 += n; 2084 p2 += n;
(...skipping 30 matching lines...) Expand all
1571 } 2115 }
1572 } 2116 }
1573 2117
1574 *p++ = POS_END; 2118 *p++ = POS_END;
1575 *pp = p; 2119 *pp = p;
1576 *pp1 = p1 + 1; 2120 *pp1 = p1 + 1;
1577 *pp2 = p2 + 1; 2121 *pp2 = p2 + 1;
1578 } 2122 }
1579 2123
1580 /* 2124 /*
1581 ** nToken==1 searches for adjacent positions.
1582 **
1583 ** This function is used to merge two position lists into one. When it is 2125 ** This function is used to merge two position lists into one. When it is
1584 ** called, *pp1 and *pp2 must both point to position lists. A position-list is 2126 ** called, *pp1 and *pp2 must both point to position lists. A position-list is
1585 ** the part of a doclist that follows each document id. For example, if a row 2127 ** the part of a doclist that follows each document id. For example, if a row
1586 ** contains: 2128 ** contains:
1587 ** 2129 **
1588 ** 'a b c'|'x y z'|'a b b a' 2130 ** 'a b c'|'x y z'|'a b b a'
1589 ** 2131 **
1590 ** Then the position list for this row for token 'b' would consist of: 2132 ** Then the position list for this row for token 'b' would consist of:
1591 ** 2133 **
1592 ** 0x02 0x01 0x02 0x03 0x03 0x00 2134 ** 0x02 0x01 0x02 0x03 0x03 0x00
1593 ** 2135 **
1594 ** When this function returns, both *pp1 and *pp2 are left pointing to the 2136 ** When this function returns, both *pp1 and *pp2 are left pointing to the
1595 ** byte following the 0x00 terminator of their respective position lists. 2137 ** byte following the 0x00 terminator of their respective position lists.
1596 ** 2138 **
1597 ** If isSaveLeft is 0, an entry is added to the output position list for 2139 ** If isSaveLeft is 0, an entry is added to the output position list for
1598 ** each position in *pp2 for which there exists one or more positions in 2140 ** each position in *pp2 for which there exists one or more positions in
1599 ** *pp1 so that (pos(*pp2)>pos(*pp1) && pos(*pp2)-pos(*pp1)<=nToken). i.e. 2141 ** *pp1 so that (pos(*pp2)>pos(*pp1) && pos(*pp2)-pos(*pp1)<=nToken). i.e.
1600 ** when the *pp1 token appears before the *pp2 token, but not more than nToken 2142 ** when the *pp1 token appears before the *pp2 token, but not more than nToken
1601 ** slots before it. 2143 ** slots before it.
2144 **
2145 ** e.g. nToken==1 searches for adjacent positions.
1602 */ 2146 */
1603 static int fts3PoslistPhraseMerge( 2147 static int fts3PoslistPhraseMerge(
1604 char **pp, /* IN/OUT: Preallocated output buffer */ 2148 char **pp, /* IN/OUT: Preallocated output buffer */
1605 int nToken, /* Maximum difference in token positions */ 2149 int nToken, /* Maximum difference in token positions */
1606 int isSaveLeft, /* Save the left position */ 2150 int isSaveLeft, /* Save the left position */
1607 int isExact, /* If *pp1 is exactly nTokens before *pp2 */ 2151 int isExact, /* If *pp1 is exactly nTokens before *pp2 */
1608 char **pp1, /* IN/OUT: Left input list */ 2152 char **pp1, /* IN/OUT: Left input list */
1609 char **pp2 /* IN/OUT: Right input list */ 2153 char **pp2 /* IN/OUT: Right input list */
1610 ){ 2154 ){
1611 char *p = (pp ? *pp : 0); 2155 char *p = *pp;
1612 char *p1 = *pp1; 2156 char *p1 = *pp1;
1613 char *p2 = *pp2; 2157 char *p2 = *pp2;
1614 int iCol1 = 0; 2158 int iCol1 = 0;
1615 int iCol2 = 0; 2159 int iCol2 = 0;
1616 2160
1617 /* Never set both isSaveLeft and isExact for the same invocation. */ 2161 /* Never set both isSaveLeft and isExact for the same invocation. */
1618 assert( isSaveLeft==0 || isExact==0 ); 2162 assert( isSaveLeft==0 || isExact==0 );
1619 2163
1620 assert( *p1!=0 && *p2!=0 ); 2164 assert( p!=0 && *p1!=0 && *p2!=0 );
1621 if( *p1==POS_COLUMN ){ 2165 if( *p1==POS_COLUMN ){
1622 p1++; 2166 p1++;
1623 p1 += sqlite3Fts3GetVarint32(p1, &iCol1); 2167 p1 += fts3GetVarint32(p1, &iCol1);
1624 } 2168 }
1625 if( *p2==POS_COLUMN ){ 2169 if( *p2==POS_COLUMN ){
1626 p2++; 2170 p2++;
1627 p2 += sqlite3Fts3GetVarint32(p2, &iCol2); 2171 p2 += fts3GetVarint32(p2, &iCol2);
1628 } 2172 }
1629 2173
1630 while( 1 ){ 2174 while( 1 ){
1631 if( iCol1==iCol2 ){ 2175 if( iCol1==iCol2 ){
1632 char *pSave = p; 2176 char *pSave = p;
1633 sqlite3_int64 iPrev = 0; 2177 sqlite3_int64 iPrev = 0;
1634 sqlite3_int64 iPos1 = 0; 2178 sqlite3_int64 iPos1 = 0;
1635 sqlite3_int64 iPos2 = 0; 2179 sqlite3_int64 iPos2 = 0;
1636 2180
1637 if( pp && iCol1 ){ 2181 if( iCol1 ){
1638 *p++ = POS_COLUMN; 2182 *p++ = POS_COLUMN;
1639 p += sqlite3Fts3PutVarint(p, iCol1); 2183 p += sqlite3Fts3PutVarint(p, iCol1);
1640 } 2184 }
1641 2185
1642 assert( *p1!=POS_END && *p1!=POS_COLUMN ); 2186 assert( *p1!=POS_END && *p1!=POS_COLUMN );
1643 assert( *p2!=POS_END && *p2!=POS_COLUMN ); 2187 assert( *p2!=POS_END && *p2!=POS_COLUMN );
1644 fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2; 2188 fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2;
1645 fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2; 2189 fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2;
1646 2190
1647 while( 1 ){ 2191 while( 1 ){
1648 if( iPos2==iPos1+nToken 2192 if( iPos2==iPos1+nToken
1649 || (isExact==0 && iPos2>iPos1 && iPos2<=iPos1+nToken) 2193 || (isExact==0 && iPos2>iPos1 && iPos2<=iPos1+nToken)
1650 ){ 2194 ){
1651 sqlite3_int64 iSave; 2195 sqlite3_int64 iSave;
1652 if( !pp ){
1653 fts3PoslistCopy(0, &p2);
1654 fts3PoslistCopy(0, &p1);
1655 *pp1 = p1;
1656 *pp2 = p2;
1657 return 1;
1658 }
1659 iSave = isSaveLeft ? iPos1 : iPos2; 2196 iSave = isSaveLeft ? iPos1 : iPos2;
1660 fts3PutDeltaVarint(&p, &iPrev, iSave+2); iPrev -= 2; 2197 fts3PutDeltaVarint(&p, &iPrev, iSave+2); iPrev -= 2;
1661 pSave = 0; 2198 pSave = 0;
2199 assert( p );
1662 } 2200 }
1663 if( (!isSaveLeft && iPos2<=(iPos1+nToken)) || iPos2<=iPos1 ){ 2201 if( (!isSaveLeft && iPos2<=(iPos1+nToken)) || iPos2<=iPos1 ){
1664 if( (*p2&0xFE)==0 ) break; 2202 if( (*p2&0xFE)==0 ) break;
1665 fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2; 2203 fts3GetDeltaVarint(&p2, &iPos2); iPos2 -= 2;
1666 }else{ 2204 }else{
1667 if( (*p1&0xFE)==0 ) break; 2205 if( (*p1&0xFE)==0 ) break;
1668 fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2; 2206 fts3GetDeltaVarint(&p1, &iPos1); iPos1 -= 2;
1669 } 2207 }
1670 } 2208 }
1671 2209
1672 if( pSave ){ 2210 if( pSave ){
1673 assert( pp && p ); 2211 assert( pp && p );
1674 p = pSave; 2212 p = pSave;
1675 } 2213 }
1676 2214
1677 fts3ColumnlistCopy(0, &p1); 2215 fts3ColumnlistCopy(0, &p1);
1678 fts3ColumnlistCopy(0, &p2); 2216 fts3ColumnlistCopy(0, &p2);
1679 assert( (*p1&0xFE)==0 && (*p2&0xFE)==0 ); 2217 assert( (*p1&0xFE)==0 && (*p2&0xFE)==0 );
1680 if( 0==*p1 || 0==*p2 ) break; 2218 if( 0==*p1 || 0==*p2 ) break;
1681 2219
1682 p1++; 2220 p1++;
1683 p1 += sqlite3Fts3GetVarint32(p1, &iCol1); 2221 p1 += fts3GetVarint32(p1, &iCol1);
1684 p2++; 2222 p2++;
1685 p2 += sqlite3Fts3GetVarint32(p2, &iCol2); 2223 p2 += fts3GetVarint32(p2, &iCol2);
1686 } 2224 }
1687 2225
1688 /* Advance pointer p1 or p2 (whichever corresponds to the smaller of 2226 /* Advance pointer p1 or p2 (whichever corresponds to the smaller of
1689 ** iCol1 and iCol2) so that it points to either the 0x00 that marks the 2227 ** iCol1 and iCol2) so that it points to either the 0x00 that marks the
1690 ** end of the position list, or the 0x01 that precedes the next 2228 ** end of the position list, or the 0x01 that precedes the next
1691 ** column-number in the position list. 2229 ** column-number in the position list.
1692 */ 2230 */
1693 else if( iCol1<iCol2 ){ 2231 else if( iCol1<iCol2 ){
1694 fts3ColumnlistCopy(0, &p1); 2232 fts3ColumnlistCopy(0, &p1);
1695 if( 0==*p1 ) break; 2233 if( 0==*p1 ) break;
1696 p1++; 2234 p1++;
1697 p1 += sqlite3Fts3GetVarint32(p1, &iCol1); 2235 p1 += fts3GetVarint32(p1, &iCol1);
1698 }else{ 2236 }else{
1699 fts3ColumnlistCopy(0, &p2); 2237 fts3ColumnlistCopy(0, &p2);
1700 if( 0==*p2 ) break; 2238 if( 0==*p2 ) break;
1701 p2++; 2239 p2++;
1702 p2 += sqlite3Fts3GetVarint32(p2, &iCol2); 2240 p2 += fts3GetVarint32(p2, &iCol2);
1703 } 2241 }
1704 } 2242 }
1705 2243
1706 fts3PoslistCopy(0, &p2); 2244 fts3PoslistCopy(0, &p2);
1707 fts3PoslistCopy(0, &p1); 2245 fts3PoslistCopy(0, &p1);
1708 *pp1 = p1; 2246 *pp1 = p1;
1709 *pp2 = p2; 2247 *pp2 = p2;
1710 if( !pp || *pp==p ){ 2248 if( *pp==p ){
1711 return 0; 2249 return 0;
1712 } 2250 }
1713 *p++ = 0x00; 2251 *p++ = 0x00;
1714 *pp = p; 2252 *pp = p;
1715 return 1; 2253 return 1;
1716 } 2254 }
1717 2255
1718 /* 2256 /*
1719 ** Merge two position-lists as required by the NEAR operator. 2257 ** Merge two position-lists as required by the NEAR operator. The argument
2258 ** position lists correspond to the left and right phrases of an expression
2259 ** like:
2260 **
2261 ** "phrase 1" NEAR "phrase number 2"
2262 **
2263 ** Position list *pp1 corresponds to the left-hand side of the NEAR
2264 ** expression and *pp2 to the right. As usual, the indexes in the position
2265 ** lists are the offsets of the last token in each phrase (tokens "1" and "2"
2266 ** in the example above).
2267 **
2268 ** The output position list - written to *pp - is a copy of *pp2 with those
2269 ** entries that are not sufficiently NEAR entries in *pp1 removed.
1720 */ 2270 */
1721 static int fts3PoslistNearMerge( 2271 static int fts3PoslistNearMerge(
1722 char **pp, /* Output buffer */ 2272 char **pp, /* Output buffer */
1723 char *aTmp, /* Temporary buffer space */ 2273 char *aTmp, /* Temporary buffer space */
1724 int nRight, /* Maximum difference in token positions */ 2274 int nRight, /* Maximum difference in token positions */
1725 int nLeft, /* Maximum difference in token positions */ 2275 int nLeft, /* Maximum difference in token positions */
1726 char **pp1, /* IN/OUT: Left input list */ 2276 char **pp1, /* IN/OUT: Left input list */
1727 char **pp2 /* IN/OUT: Right input list */ 2277 char **pp2 /* IN/OUT: Right input list */
1728 ){ 2278 ){
1729 char *p1 = *pp1; 2279 char *p1 = *pp1;
1730 char *p2 = *pp2; 2280 char *p2 = *pp2;
1731 2281
1732 if( !pp ){ 2282 char *pTmp1 = aTmp;
1733 if( fts3PoslistPhraseMerge(0, nRight, 0, 0, pp1, pp2) ) return 1; 2283 char *pTmp2;
1734 *pp1 = p1; 2284 char *aTmp2;
1735 *pp2 = p2; 2285 int res = 1;
1736 return fts3PoslistPhraseMerge(0, nLeft, 0, 0, pp2, pp1); 2286
2287 fts3PoslistPhraseMerge(&pTmp1, nRight, 0, 0, pp1, pp2);
2288 aTmp2 = pTmp2 = pTmp1;
2289 *pp1 = p1;
2290 *pp2 = p2;
2291 fts3PoslistPhraseMerge(&pTmp2, nLeft, 1, 0, pp2, pp1);
2292 if( pTmp1!=aTmp && pTmp2!=aTmp2 ){
2293 fts3PoslistMerge(pp, &aTmp, &aTmp2);
2294 }else if( pTmp1!=aTmp ){
2295 fts3PoslistCopy(pp, &aTmp);
2296 }else if( pTmp2!=aTmp2 ){
2297 fts3PoslistCopy(pp, &aTmp2);
1737 }else{ 2298 }else{
1738 char *pTmp1 = aTmp; 2299 res = 0;
1739 char *pTmp2; 2300 }
1740 char *aTmp2; 2301
1741 int res = 1; 2302 return res;
1742 2303 }
1743 fts3PoslistPhraseMerge(&pTmp1, nRight, 0, 0, pp1, pp2); 2304
1744 aTmp2 = pTmp2 = pTmp1; 2305 /*
1745 *pp1 = p1; 2306 ** An instance of this function is used to merge together the (potentially
1746 *pp2 = p2; 2307 ** large number of) doclists for each term that matches a prefix query.
1747 fts3PoslistPhraseMerge(&pTmp2, nLeft, 1, 0, pp2, pp1); 2308 ** See function fts3TermSelectMerge() for details.
1748 if( pTmp1!=aTmp && pTmp2!=aTmp2 ){ 2309 */
1749 fts3PoslistMerge(pp, &aTmp, &aTmp2); 2310 typedef struct TermSelect TermSelect;
1750 }else if( pTmp1!=aTmp ){ 2311 struct TermSelect {
1751 fts3PoslistCopy(pp, &aTmp); 2312 char *aaOutput[16]; /* Malloc'd output buffers */
1752 }else if( pTmp2!=aTmp2 ){ 2313 int anOutput[16]; /* Size each output buffer in bytes */
1753 fts3PoslistCopy(pp, &aTmp2); 2314 };
2315
2316 /*
2317 ** This function is used to read a single varint from a buffer. Parameter
2318 ** pEnd points 1 byte past the end of the buffer. When this function is
2319 ** called, if *pp points to pEnd or greater, then the end of the buffer
2320 ** has been reached. In this case *pp is set to 0 and the function returns.
2321 **
2322 ** If *pp does not point to or past pEnd, then a single varint is read
2323 ** from *pp. *pp is then set to point 1 byte past the end of the read varint.
2324 **
2325 ** If bDescIdx is false, the value read is added to *pVal before returning.
2326 ** If it is true, the value read is subtracted from *pVal before this
2327 ** function returns.
2328 */
2329 static void fts3GetDeltaVarint3(
2330 char **pp, /* IN/OUT: Point to read varint from */
2331 char *pEnd, /* End of buffer */
2332 int bDescIdx, /* True if docids are descending */
2333 sqlite3_int64 *pVal /* IN/OUT: Integer value */
2334 ){
2335 if( *pp>=pEnd ){
2336 *pp = 0;
2337 }else{
2338 sqlite3_int64 iVal;
2339 *pp += sqlite3Fts3GetVarint(*pp, &iVal);
2340 if( bDescIdx ){
2341 *pVal -= iVal;
1754 }else{ 2342 }else{
1755 res = 0; 2343 *pVal += iVal;
1756 } 2344 }
1757 2345 }
1758 return res; 2346 }
1759 } 2347
1760 } 2348 /*
1761 2349 ** This function is used to write a single varint to a buffer. The varint
1762 /* 2350 ** is written to *pp. Before returning, *pp is set to point 1 byte past the
1763 ** Values that may be used as the first parameter to fts3DoclistMerge(). 2351 ** end of the value written.
1764 */ 2352 **
1765 #define MERGE_NOT 2 /* D + D -> D */ 2353 ** If *pbFirst is zero when this function is called, the value written to
1766 #define MERGE_AND 3 /* D + D -> D */ 2354 ** the buffer is that of parameter iVal.
1767 #define MERGE_OR 4 /* D + D -> D */ 2355 **
1768 #define MERGE_POS_OR 5 /* P + P -> P */ 2356 ** If *pbFirst is non-zero when this function is called, then the value
1769 #define MERGE_PHRASE 6 /* P + P -> D */ 2357 ** written is either (iVal-*piPrev) (if bDescIdx is zero) or (*piPrev-iVal)
1770 #define MERGE_POS_PHRASE 7 /* P + P -> P */ 2358 ** (if bDescIdx is non-zero).
1771 #define MERGE_NEAR 8 /* P + P -> D */ 2359 **
1772 #define MERGE_POS_NEAR 9 /* P + P -> P */ 2360 ** Before returning, this function always sets *pbFirst to 1 and *piPrev
1773 2361 ** to the value of parameter iVal.
1774 /* 2362 */
1775 ** Merge the two doclists passed in buffer a1 (size n1 bytes) and a2 2363 static void fts3PutDeltaVarint3(
1776 ** (size n2 bytes). The output is written to pre-allocated buffer aBuffer, 2364 char **pp, /* IN/OUT: Output pointer */
1777 ** which is guaranteed to be large enough to hold the results. The number 2365 int bDescIdx, /* True for descending docids */
1778 ** of bytes written to aBuffer is stored in *pnBuffer before returning. 2366 sqlite3_int64 *piPrev, /* IN/OUT: Previous value written to list */
1779 ** 2367 int *pbFirst, /* IN/OUT: True after first int written */
1780 ** If successful, SQLITE_OK is returned. Otherwise, if a malloc error 2368 sqlite3_int64 iVal /* Write this value to the list */
1781 ** occurs while allocating a temporary buffer as part of the merge operation, 2369 ){
1782 ** SQLITE_NOMEM is returned. 2370 sqlite3_int64 iWrite;
1783 */ 2371 if( bDescIdx==0 || *pbFirst==0 ){
1784 static int fts3DoclistMerge( 2372 iWrite = iVal - *piPrev;
1785 int mergetype, /* One of the MERGE_XXX constants */ 2373 }else{
1786 int nParam1, /* Used by MERGE_NEAR and MERGE_POS_NEAR */ 2374 iWrite = *piPrev - iVal;
1787 int nParam2, /* Used by MERGE_NEAR and MERGE_POS_NEAR */ 2375 }
1788 char *aBuffer, /* Pre-allocated output buffer */ 2376 assert( *pbFirst || *piPrev==0 );
1789 int *pnBuffer, /* OUT: Bytes written to aBuffer */ 2377 assert( *pbFirst==0 || iWrite>0 );
1790 char *a1, /* Buffer containing first doclist */ 2378 *pp += sqlite3Fts3PutVarint(*pp, iWrite);
1791 int n1, /* Size of buffer a1 */ 2379 *piPrev = iVal;
1792 char *a2, /* Buffer containing second doclist */ 2380 *pbFirst = 1;
1793 int n2, /* Size of buffer a2 */ 2381 }
1794 int *pnDoc /* OUT: Number of docids in output */ 2382
2383
2384 /*
2385 ** This macro is used by various functions that merge doclists. The two
2386 ** arguments are 64-bit docid values. If the value of the stack variable
2387 ** bDescDoclist is 0 when this macro is invoked, then it returns (i1-i2).
2388 ** Otherwise, (i2-i1).
2389 **
2390 ** Using this makes it easier to write code that can merge doclists that are
2391 ** sorted in either ascending or descending order.
2392 */
2393 #define DOCID_CMP(i1, i2) ((bDescDoclist?-1:1) * (i1-i2))
2394
2395 /*
2396 ** This function does an "OR" merge of two doclists (output contains all
2397 ** positions contained in either argument doclist). If the docids in the
2398 ** input doclists are sorted in ascending order, parameter bDescDoclist
2399 ** should be false. If they are sorted in ascending order, it should be
2400 ** passed a non-zero value.
2401 **
2402 ** If no error occurs, *paOut is set to point at an sqlite3_malloc'd buffer
2403 ** containing the output doclist and SQLITE_OK is returned. In this case
2404 ** *pnOut is set to the number of bytes in the output doclist.
2405 **
2406 ** If an error occurs, an SQLite error code is returned. The output values
2407 ** are undefined in this case.
2408 */
2409 static int fts3DoclistOrMerge(
2410 int bDescDoclist, /* True if arguments are desc */
2411 char *a1, int n1, /* First doclist */
2412 char *a2, int n2, /* Second doclist */
2413 char **paOut, int *pnOut /* OUT: Malloc'd doclist */
1795 ){ 2414 ){
1796 sqlite3_int64 i1 = 0; 2415 sqlite3_int64 i1 = 0;
1797 sqlite3_int64 i2 = 0; 2416 sqlite3_int64 i2 = 0;
1798 sqlite3_int64 iPrev = 0; 2417 sqlite3_int64 iPrev = 0;
1799 2418 char *pEnd1 = &a1[n1];
1800 char *p = aBuffer; 2419 char *pEnd2 = &a2[n2];
1801 char *p1 = a1; 2420 char *p1 = a1;
1802 char *p2 = a2; 2421 char *p2 = a2;
1803 char *pEnd1 = &a1[n1]; 2422 char *p;
1804 char *pEnd2 = &a2[n2]; 2423 char *aOut;
1805 int nDoc = 0; 2424 int bFirstOut = 0;
1806 2425
1807 assert( mergetype==MERGE_OR || mergetype==MERGE_POS_OR 2426 *paOut = 0;
1808 || mergetype==MERGE_AND || mergetype==MERGE_NOT 2427 *pnOut = 0;
1809 || mergetype==MERGE_PHRASE || mergetype==MERGE_POS_PHRASE 2428
1810 || mergetype==MERGE_NEAR || mergetype==MERGE_POS_NEAR 2429 /* Allocate space for the output. Both the input and output doclists
1811 ); 2430 ** are delta encoded. If they are in ascending order (bDescDoclist==0),
1812 2431 ** then the first docid in each list is simply encoded as a varint. For
1813 if( !aBuffer ){ 2432 ** each subsequent docid, the varint stored is the difference between the
1814 *pnBuffer = 0; 2433 ** current and previous docid (a positive number - since the list is in
1815 return SQLITE_NOMEM; 2434 ** ascending order).
1816 } 2435 **
1817 2436 ** The first docid written to the output is therefore encoded using the
1818 /* Read the first docid from each doclist */ 2437 ** same number of bytes as it is in whichever of the input lists it is
1819 fts3GetDeltaVarint2(&p1, pEnd1, &i1); 2438 ** read from. And each subsequent docid read from the same input list
1820 fts3GetDeltaVarint2(&p2, pEnd2, &i2); 2439 ** consumes either the same or less bytes as it did in the input (since
1821 2440 ** the difference between it and the previous value in the output must
1822 switch( mergetype ){ 2441 ** be a positive value less than or equal to the delta value read from
1823 case MERGE_OR: 2442 ** the input list). The same argument applies to all but the first docid
1824 case MERGE_POS_OR: 2443 ** read from the 'other' list. And to the contents of all position lists
1825 while( p1 || p2 ){ 2444 ** that will be copied and merged from the input to the output.
1826 if( p2 && p1 && i1==i2 ){ 2445 **
1827 fts3PutDeltaVarint(&p, &iPrev, i1); 2446 ** However, if the first docid copied to the output is a negative number,
1828 if( mergetype==MERGE_POS_OR ) fts3PoslistMerge(&p, &p1, &p2); 2447 ** then the encoding of the first docid from the 'other' input list may
1829 fts3GetDeltaVarint2(&p1, pEnd1, &i1); 2448 ** be larger in the output than it was in the input (since the delta value
1830 fts3GetDeltaVarint2(&p2, pEnd2, &i2); 2449 ** may be a larger positive integer than the actual docid).
1831 }else if( !p2 || (p1 && i1<i2) ){ 2450 **
1832 fts3PutDeltaVarint(&p, &iPrev, i1); 2451 ** The space required to store the output is therefore the sum of the
1833 if( mergetype==MERGE_POS_OR ) fts3PoslistCopy(&p, &p1); 2452 ** sizes of the two inputs, plus enough space for exactly one of the input
1834 fts3GetDeltaVarint2(&p1, pEnd1, &i1); 2453 ** docids to grow.
1835 }else{ 2454 **
1836 fts3PutDeltaVarint(&p, &iPrev, i2); 2455 ** A symetric argument may be made if the doclists are in descending
1837 if( mergetype==MERGE_POS_OR ) fts3PoslistCopy(&p, &p2); 2456 ** order.
1838 fts3GetDeltaVarint2(&p2, pEnd2, &i2); 2457 */
1839 } 2458 aOut = sqlite3_malloc(n1+n2+FTS3_VARINT_MAX-1);
2459 if( !aOut ) return SQLITE_NOMEM;
2460
2461 p = aOut;
2462 fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
2463 fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
2464 while( p1 || p2 ){
2465 sqlite3_int64 iDiff = DOCID_CMP(i1, i2);
2466
2467 if( p2 && p1 && iDiff==0 ){
2468 fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
2469 fts3PoslistMerge(&p, &p1, &p2);
2470 fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
2471 fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
2472 }else if( !p2 || (p1 && iDiff<0) ){
2473 fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
2474 fts3PoslistCopy(&p, &p1);
2475 fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
2476 }else{
2477 fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i2);
2478 fts3PoslistCopy(&p, &p2);
2479 fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
2480 }
2481 }
2482
2483 *paOut = aOut;
2484 *pnOut = (int)(p-aOut);
2485 assert( *pnOut<=n1+n2+FTS3_VARINT_MAX-1 );
2486 return SQLITE_OK;
2487 }
2488
2489 /*
2490 ** This function does a "phrase" merge of two doclists. In a phrase merge,
2491 ** the output contains a copy of each position from the right-hand input
2492 ** doclist for which there is a position in the left-hand input doclist
2493 ** exactly nDist tokens before it.
2494 **
2495 ** If the docids in the input doclists are sorted in ascending order,
2496 ** parameter bDescDoclist should be false. If they are sorted in ascending
2497 ** order, it should be passed a non-zero value.
2498 **
2499 ** The right-hand input doclist is overwritten by this function.
2500 */
2501 static void fts3DoclistPhraseMerge(
2502 int bDescDoclist, /* True if arguments are desc */
2503 int nDist, /* Distance from left to right (1=adjacent) */
2504 char *aLeft, int nLeft, /* Left doclist */
2505 char *aRight, int *pnRight /* IN/OUT: Right/output doclist */
2506 ){
2507 sqlite3_int64 i1 = 0;
2508 sqlite3_int64 i2 = 0;
2509 sqlite3_int64 iPrev = 0;
2510 char *pEnd1 = &aLeft[nLeft];
2511 char *pEnd2 = &aRight[*pnRight];
2512 char *p1 = aLeft;
2513 char *p2 = aRight;
2514 char *p;
2515 int bFirstOut = 0;
2516 char *aOut = aRight;
2517
2518 assert( nDist>0 );
2519
2520 p = aOut;
2521 fts3GetDeltaVarint3(&p1, pEnd1, 0, &i1);
2522 fts3GetDeltaVarint3(&p2, pEnd2, 0, &i2);
2523
2524 while( p1 && p2 ){
2525 sqlite3_int64 iDiff = DOCID_CMP(i1, i2);
2526 if( iDiff==0 ){
2527 char *pSave = p;
2528 sqlite3_int64 iPrevSave = iPrev;
2529 int bFirstOutSave = bFirstOut;
2530
2531 fts3PutDeltaVarint3(&p, bDescDoclist, &iPrev, &bFirstOut, i1);
2532 if( 0==fts3PoslistPhraseMerge(&p, nDist, 0, 1, &p1, &p2) ){
2533 p = pSave;
2534 iPrev = iPrevSave;
2535 bFirstOut = bFirstOutSave;
1840 } 2536 }
1841 break; 2537 fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
1842 2538 fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
1843 case MERGE_AND: 2539 }else if( iDiff<0 ){
1844 while( p1 && p2 ){ 2540 fts3PoslistCopy(0, &p1);
1845 if( i1==i2 ){ 2541 fts3GetDeltaVarint3(&p1, pEnd1, bDescDoclist, &i1);
1846 fts3PutDeltaVarint(&p, &iPrev, i1); 2542 }else{
1847 fts3GetDeltaVarint2(&p1, pEnd1, &i1); 2543 fts3PoslistCopy(0, &p2);
1848 fts3GetDeltaVarint2(&p2, pEnd2, &i2); 2544 fts3GetDeltaVarint3(&p2, pEnd2, bDescDoclist, &i2);
1849 nDoc++; 2545 }
1850 }else if( i1<i2 ){ 2546 }
1851 fts3GetDeltaVarint2(&p1, pEnd1, &i1); 2547
1852 }else{ 2548 *pnRight = (int)(p - aOut);
1853 fts3GetDeltaVarint2(&p2, pEnd2, &i2); 2549 }
1854 } 2550
2551 /*
2552 ** Argument pList points to a position list nList bytes in size. This
2553 ** function checks to see if the position list contains any entries for
2554 ** a token in position 0 (of any column). If so, it writes argument iDelta
2555 ** to the output buffer pOut, followed by a position list consisting only
2556 ** of the entries from pList at position 0, and terminated by an 0x00 byte.
2557 ** The value returned is the number of bytes written to pOut (if any).
2558 */
2559 int sqlite3Fts3FirstFilter(
2560 sqlite3_int64 iDelta, /* Varint that may be written to pOut */
2561 char *pList, /* Position list (no 0x00 term) */
2562 int nList, /* Size of pList in bytes */
2563 char *pOut /* Write output here */
2564 ){
2565 int nOut = 0;
2566 int bWritten = 0; /* True once iDelta has been written */
2567 char *p = pList;
2568 char *pEnd = &pList[nList];
2569
2570 if( *p!=0x01 ){
2571 if( *p==0x02 ){
2572 nOut += sqlite3Fts3PutVarint(&pOut[nOut], iDelta);
2573 pOut[nOut++] = 0x02;
2574 bWritten = 1;
2575 }
2576 fts3ColumnlistCopy(0, &p);
2577 }
2578
2579 while( p<pEnd && *p==0x01 ){
2580 sqlite3_int64 iCol;
2581 p++;
2582 p += sqlite3Fts3GetVarint(p, &iCol);
2583 if( *p==0x02 ){
2584 if( bWritten==0 ){
2585 nOut += sqlite3Fts3PutVarint(&pOut[nOut], iDelta);
2586 bWritten = 1;
1855 } 2587 }
1856 break; 2588 pOut[nOut++] = 0x01;
1857 2589 nOut += sqlite3Fts3PutVarint(&pOut[nOut], iCol);
1858 case MERGE_NOT: 2590 pOut[nOut++] = 0x02;
1859 while( p1 ){ 2591 }
1860 if( p2 && i1==i2 ){ 2592 fts3ColumnlistCopy(0, &p);
1861 fts3GetDeltaVarint2(&p1, pEnd1, &i1); 2593 }
1862 fts3GetDeltaVarint2(&p2, pEnd2, &i2); 2594 if( bWritten ){
1863 }else if( !p2 || i1<i2 ){ 2595 pOut[nOut++] = 0x00;
1864 fts3PutDeltaVarint(&p, &iPrev, i1); 2596 }
1865 fts3GetDeltaVarint2(&p1, pEnd1, &i1); 2597
1866 }else{ 2598 return nOut;
1867 fts3GetDeltaVarint2(&p2, pEnd2, &i2); 2599 }
1868 } 2600
1869 }
1870 break;
1871
1872 case MERGE_POS_PHRASE:
1873 case MERGE_PHRASE: {
1874 char **ppPos = (mergetype==MERGE_PHRASE ? 0 : &p);
1875 while( p1 && p2 ){
1876 if( i1==i2 ){
1877 char *pSave = p;
1878 sqlite3_int64 iPrevSave = iPrev;
1879 fts3PutDeltaVarint(&p, &iPrev, i1);
1880 if( 0==fts3PoslistPhraseMerge(ppPos, nParam1, 0, 1, &p1, &p2) ){
1881 p = pSave;
1882 iPrev = iPrevSave;
1883 }else{
1884 nDoc++;
1885 }
1886 fts3GetDeltaVarint2(&p1, pEnd1, &i1);
1887 fts3GetDeltaVarint2(&p2, pEnd2, &i2);
1888 }else if( i1<i2 ){
1889 fts3PoslistCopy(0, &p1);
1890 fts3GetDeltaVarint2(&p1, pEnd1, &i1);
1891 }else{
1892 fts3PoslistCopy(0, &p2);
1893 fts3GetDeltaVarint2(&p2, pEnd2, &i2);
1894 }
1895 }
1896 break;
1897 }
1898
1899 default: assert( mergetype==MERGE_POS_NEAR || mergetype==MERGE_NEAR ); {
1900 char *aTmp = 0;
1901 char **ppPos = 0;
1902
1903 if( mergetype==MERGE_POS_NEAR ){
1904 ppPos = &p;
1905 aTmp = sqlite3_malloc(2*(n1+n2+1));
1906 if( !aTmp ){
1907 return SQLITE_NOMEM;
1908 }
1909 }
1910
1911 while( p1 && p2 ){
1912 if( i1==i2 ){
1913 char *pSave = p;
1914 sqlite3_int64 iPrevSave = iPrev;
1915 fts3PutDeltaVarint(&p, &iPrev, i1);
1916
1917 if( !fts3PoslistNearMerge(ppPos, aTmp, nParam1, nParam2, &p1, &p2) ){
1918 iPrev = iPrevSave;
1919 p = pSave;
1920 }
1921
1922 fts3GetDeltaVarint2(&p1, pEnd1, &i1);
1923 fts3GetDeltaVarint2(&p2, pEnd2, &i2);
1924 }else if( i1<i2 ){
1925 fts3PoslistCopy(0, &p1);
1926 fts3GetDeltaVarint2(&p1, pEnd1, &i1);
1927 }else{
1928 fts3PoslistCopy(0, &p2);
1929 fts3GetDeltaVarint2(&p2, pEnd2, &i2);
1930 }
1931 }
1932 sqlite3_free(aTmp);
1933 break;
1934 }
1935 }
1936
1937 if( pnDoc ) *pnDoc = nDoc;
1938 *pnBuffer = (int)(p-aBuffer);
1939 return SQLITE_OK;
1940 }
1941
1942 /*
1943 ** A pointer to an instance of this structure is used as the context
1944 ** argument to sqlite3Fts3SegReaderIterate()
1945 */
1946 typedef struct TermSelect TermSelect;
1947 struct TermSelect {
1948 int isReqPos;
1949 char *aaOutput[16]; /* Malloc'd output buffer */
1950 int anOutput[16]; /* Size of output in bytes */
1951 };
1952 2601
1953 /* 2602 /*
1954 ** Merge all doclists in the TermSelect.aaOutput[] array into a single 2603 ** Merge all doclists in the TermSelect.aaOutput[] array into a single
1955 ** doclist stored in TermSelect.aaOutput[0]. If successful, delete all 2604 ** doclist stored in TermSelect.aaOutput[0]. If successful, delete all
1956 ** other doclists (except the aaOutput[0] one) and return SQLITE_OK. 2605 ** other doclists (except the aaOutput[0] one) and return SQLITE_OK.
1957 ** 2606 **
1958 ** If an OOM error occurs, return SQLITE_NOMEM. In this case it is 2607 ** If an OOM error occurs, return SQLITE_NOMEM. In this case it is
1959 ** the responsibility of the caller to free any doclists left in the 2608 ** the responsibility of the caller to free any doclists left in the
1960 ** TermSelect.aaOutput[] array. 2609 ** TermSelect.aaOutput[] array.
1961 */ 2610 */
1962 static int fts3TermSelectMerge(TermSelect *pTS){ 2611 static int fts3TermSelectFinishMerge(Fts3Table *p, TermSelect *pTS){
1963 int mergetype = (pTS->isReqPos ? MERGE_POS_OR : MERGE_OR);
1964 char *aOut = 0; 2612 char *aOut = 0;
1965 int nOut = 0; 2613 int nOut = 0;
1966 int i; 2614 int i;
1967 2615
1968 /* Loop through the doclists in the aaOutput[] array. Merge them all 2616 /* Loop through the doclists in the aaOutput[] array. Merge them all
1969 ** into a single doclist. 2617 ** into a single doclist.
1970 */ 2618 */
1971 for(i=0; i<SizeofArray(pTS->aaOutput); i++){ 2619 for(i=0; i<SizeofArray(pTS->aaOutput); i++){
1972 if( pTS->aaOutput[i] ){ 2620 if( pTS->aaOutput[i] ){
1973 if( !aOut ){ 2621 if( !aOut ){
1974 aOut = pTS->aaOutput[i]; 2622 aOut = pTS->aaOutput[i];
1975 nOut = pTS->anOutput[i]; 2623 nOut = pTS->anOutput[i];
1976 pTS->aaOutput[i] = 0; 2624 pTS->aaOutput[i] = 0;
1977 }else{ 2625 }else{
1978 int nNew = nOut + pTS->anOutput[i]; 2626 int nNew;
1979 char *aNew = sqlite3_malloc(nNew); 2627 char *aNew;
1980 if( !aNew ){ 2628
2629 int rc = fts3DoclistOrMerge(p->bDescIdx,
2630 pTS->aaOutput[i], pTS->anOutput[i], aOut, nOut, &aNew, &nNew
2631 );
2632 if( rc!=SQLITE_OK ){
1981 sqlite3_free(aOut); 2633 sqlite3_free(aOut);
1982 return SQLITE_NOMEM; 2634 return rc;
1983 } 2635 }
1984 fts3DoclistMerge(mergetype, 0, 0, 2636
1985 aNew, &nNew, pTS->aaOutput[i], pTS->anOutput[i], aOut, nOut, 0
1986 );
1987 sqlite3_free(pTS->aaOutput[i]); 2637 sqlite3_free(pTS->aaOutput[i]);
1988 sqlite3_free(aOut); 2638 sqlite3_free(aOut);
1989 pTS->aaOutput[i] = 0; 2639 pTS->aaOutput[i] = 0;
1990 aOut = aNew; 2640 aOut = aNew;
1991 nOut = nNew; 2641 nOut = nNew;
1992 } 2642 }
1993 } 2643 }
1994 } 2644 }
1995 2645
1996 pTS->aaOutput[0] = aOut; 2646 pTS->aaOutput[0] = aOut;
1997 pTS->anOutput[0] = nOut; 2647 pTS->anOutput[0] = nOut;
1998 return SQLITE_OK; 2648 return SQLITE_OK;
1999 } 2649 }
2000 2650
2001 /* 2651 /*
2002 ** This function is used as the sqlite3Fts3SegReaderIterate() callback when 2652 ** Merge the doclist aDoclist/nDoclist into the TermSelect object passed
2003 ** querying the full-text index for a doclist associated with a term or 2653 ** as the first argument. The merge is an "OR" merge (see function
2004 ** term-prefix. 2654 ** fts3DoclistOrMerge() for details).
2655 **
2656 ** This function is called with the doclist for each term that matches
2657 ** a queried prefix. It merges all these doclists into one, the doclist
2658 ** for the specified prefix. Since there can be a very large number of
2659 ** doclists to merge, the merging is done pair-wise using the TermSelect
2660 ** object.
2661 **
2662 ** This function returns SQLITE_OK if the merge is successful, or an
2663 ** SQLite error code (SQLITE_NOMEM) if an error occurs.
2005 */ 2664 */
2006 static int fts3TermSelectCb( 2665 static int fts3TermSelectMerge(
2007 Fts3Table *p, /* Virtual table object */ 2666 Fts3Table *p, /* FTS table handle */
2008 void *pContext, /* Pointer to TermSelect structure */ 2667 TermSelect *pTS, /* TermSelect object to merge into */
2009 char *zTerm, 2668 char *aDoclist, /* Pointer to doclist */
2010 int nTerm, 2669 int nDoclist /* Size of aDoclist in bytes */
2011 char *aDoclist,
2012 int nDoclist
2013 ){ 2670 ){
2014 TermSelect *pTS = (TermSelect *)pContext;
2015
2016 UNUSED_PARAMETER(p);
2017 UNUSED_PARAMETER(zTerm);
2018 UNUSED_PARAMETER(nTerm);
2019
2020 if( pTS->aaOutput[0]==0 ){ 2671 if( pTS->aaOutput[0]==0 ){
2021 /* If this is the first term selected, copy the doclist to the output 2672 /* If this is the first term selected, copy the doclist to the output
2022 ** buffer using memcpy(). TODO: Add a way to transfer control of the 2673 ** buffer using memcpy(). */
2023 ** aDoclist buffer from the caller so as to avoid the memcpy().
2024 */
2025 pTS->aaOutput[0] = sqlite3_malloc(nDoclist); 2674 pTS->aaOutput[0] = sqlite3_malloc(nDoclist);
2026 pTS->anOutput[0] = nDoclist; 2675 pTS->anOutput[0] = nDoclist;
2027 if( pTS->aaOutput[0] ){ 2676 if( pTS->aaOutput[0] ){
2028 memcpy(pTS->aaOutput[0], aDoclist, nDoclist); 2677 memcpy(pTS->aaOutput[0], aDoclist, nDoclist);
2029 }else{ 2678 }else{
2030 return SQLITE_NOMEM; 2679 return SQLITE_NOMEM;
2031 } 2680 }
2032 }else{ 2681 }else{
2033 int mergetype = (pTS->isReqPos ? MERGE_POS_OR : MERGE_OR);
2034 char *aMerge = aDoclist; 2682 char *aMerge = aDoclist;
2035 int nMerge = nDoclist; 2683 int nMerge = nDoclist;
2036 int iOut; 2684 int iOut;
2037 2685
2038 for(iOut=0; iOut<SizeofArray(pTS->aaOutput); iOut++){ 2686 for(iOut=0; iOut<SizeofArray(pTS->aaOutput); iOut++){
2039 char *aNew;
2040 int nNew;
2041 if( pTS->aaOutput[iOut]==0 ){ 2687 if( pTS->aaOutput[iOut]==0 ){
2042 assert( iOut>0 ); 2688 assert( iOut>0 );
2043 pTS->aaOutput[iOut] = aMerge; 2689 pTS->aaOutput[iOut] = aMerge;
2044 pTS->anOutput[iOut] = nMerge; 2690 pTS->anOutput[iOut] = nMerge;
2045 break; 2691 break;
2046 } 2692 }else{
2693 char *aNew;
2694 int nNew;
2047 2695
2048 nNew = nMerge + pTS->anOutput[iOut]; 2696 int rc = fts3DoclistOrMerge(p->bDescIdx, aMerge, nMerge,
2049 aNew = sqlite3_malloc(nNew); 2697 pTS->aaOutput[iOut], pTS->anOutput[iOut], &aNew, &nNew
2050 if( !aNew ){ 2698 );
2051 if( aMerge!=aDoclist ){ 2699 if( rc!=SQLITE_OK ){
2052 sqlite3_free(aMerge); 2700 if( aMerge!=aDoclist ) sqlite3_free(aMerge);
2701 return rc;
2053 } 2702 }
2054 return SQLITE_NOMEM;
2055 }
2056 fts3DoclistMerge(mergetype, 0, 0, aNew, &nNew,
2057 pTS->aaOutput[iOut], pTS->anOutput[iOut], aMerge, nMerge, 0
2058 );
2059 2703
2060 if( iOut>0 ) sqlite3_free(aMerge); 2704 if( aMerge!=aDoclist ) sqlite3_free(aMerge);
2061 sqlite3_free(pTS->aaOutput[iOut]); 2705 sqlite3_free(pTS->aaOutput[iOut]);
2062 pTS->aaOutput[iOut] = 0; 2706 pTS->aaOutput[iOut] = 0;
2063 2707
2064 aMerge = aNew; 2708 aMerge = aNew;
2065 nMerge = nNew; 2709 nMerge = nNew;
2066 if( (iOut+1)==SizeofArray(pTS->aaOutput) ){ 2710 if( (iOut+1)==SizeofArray(pTS->aaOutput) ){
2067 pTS->aaOutput[iOut] = aMerge; 2711 pTS->aaOutput[iOut] = aMerge;
2068 pTS->anOutput[iOut] = nMerge; 2712 pTS->anOutput[iOut] = nMerge;
2713 }
2069 } 2714 }
2070 } 2715 }
2071 } 2716 }
2072 return SQLITE_OK; 2717 return SQLITE_OK;
2073 } 2718 }
2074 2719
2075 static int fts3DeferredTermSelect( 2720 /*
2076 Fts3DeferredToken *pToken, /* Phrase token */ 2721 ** Append SegReader object pNew to the end of the pCsr->apSegment[] array.
2077 int isTermPos, /* True to include positions */ 2722 */
2078 int *pnOut, /* OUT: Size of list */ 2723 static int fts3SegReaderCursorAppend(
2079 char **ppOut /* OUT: Body of list */ 2724 Fts3MultiSegReader *pCsr,
2725 Fts3SegReader *pNew
2080 ){ 2726 ){
2081 char *aSource; 2727 if( (pCsr->nSegment%16)==0 ){
2082 int nSource; 2728 Fts3SegReader **apNew;
2083 2729 int nByte = (pCsr->nSegment + 16)*sizeof(Fts3SegReader*);
2084 aSource = sqlite3Fts3DeferredDoclist(pToken, &nSource); 2730 apNew = (Fts3SegReader **)sqlite3_realloc(pCsr->apSegment, nByte);
2085 if( !aSource ){ 2731 if( !apNew ){
2086 *pnOut = 0; 2732 sqlite3Fts3SegReaderFree(pNew);
2087 *ppOut = 0; 2733 return SQLITE_NOMEM;
2088 }else if( isTermPos ){ 2734 }
2089 *ppOut = sqlite3_malloc(nSource); 2735 pCsr->apSegment = apNew;
2090 if( !*ppOut ) return SQLITE_NOMEM;
2091 memcpy(*ppOut, aSource, nSource);
2092 *pnOut = nSource;
2093 }else{
2094 sqlite3_int64 docid;
2095 *pnOut = sqlite3Fts3GetVarint(aSource, &docid);
2096 *ppOut = sqlite3_malloc(*pnOut);
2097 if( !*ppOut ) return SQLITE_NOMEM;
2098 sqlite3Fts3PutVarint(*ppOut, docid);
2099 } 2736 }
2100 2737 pCsr->apSegment[pCsr->nSegment++] = pNew;
2101 return SQLITE_OK; 2738 return SQLITE_OK;
2102 } 2739 }
2103 2740
2104 int sqlite3Fts3SegReaderCursor( 2741 /*
2742 ** Add seg-reader objects to the Fts3MultiSegReader object passed as the
2743 ** 8th argument.
2744 **
2745 ** This function returns SQLITE_OK if successful, or an SQLite error code
2746 ** otherwise.
2747 */
2748 static int fts3SegReaderCursor(
2105 Fts3Table *p, /* FTS3 table handle */ 2749 Fts3Table *p, /* FTS3 table handle */
2750 int iLangid, /* Language id */
2751 int iIndex, /* Index to search (from 0 to p->nIndex-1) */
2106 int iLevel, /* Level of segments to scan */ 2752 int iLevel, /* Level of segments to scan */
2107 const char *zTerm, /* Term to query for */ 2753 const char *zTerm, /* Term to query for */
2108 int nTerm, /* Size of zTerm in bytes */ 2754 int nTerm, /* Size of zTerm in bytes */
2109 int isPrefix, /* True for a prefix search */ 2755 int isPrefix, /* True for a prefix search */
2110 int isScan, /* True to scan from zTerm to EOF */ 2756 int isScan, /* True to scan from zTerm to EOF */
2111 Fts3SegReaderCursor *pCsr /* Cursor object to populate */ 2757 Fts3MultiSegReader *pCsr /* Cursor object to populate */
2112 ){ 2758 ){
2113 int rc = SQLITE_OK; 2759 int rc = SQLITE_OK; /* Error code */
2114 int rc2; 2760 sqlite3_stmt *pStmt = 0; /* Statement to iterate through segments */
2115 int iAge = 0; 2761 int rc2; /* Result of sqlite3_reset() */
2116 sqlite3_stmt *pStmt = 0;
2117 Fts3SegReader *pPending = 0;
2118 2762
2119 assert( iLevel==FTS3_SEGCURSOR_ALL 2763 /* If iLevel is less than 0 and this is not a scan, include a seg-reader
2120 || iLevel==FTS3_SEGCURSOR_PENDING 2764 ** for the pending-terms. If this is a scan, then this call must be being
2121 || iLevel>=0 2765 ** made by an fts4aux module, not an FTS table. In this case calling
2122 ); 2766 ** Fts3SegReaderPending might segfault, as the data structures used by
2123 assert( FTS3_SEGCURSOR_PENDING<0 ); 2767 ** fts4aux are not completely populated. So it's easiest to filter these
2124 assert( FTS3_SEGCURSOR_ALL<0 ); 2768 ** calls out here. */
2125 assert( iLevel==FTS3_SEGCURSOR_ALL || (zTerm==0 && isPrefix==1) ); 2769 if( iLevel<0 && p->aIndex ){
2126 assert( isPrefix==0 || isScan==0 ); 2770 Fts3SegReader *pSeg = 0;
2127 2771 rc = sqlite3Fts3SegReaderPending(p, iIndex, zTerm, nTerm, isPrefix, &pSeg);
2128 2772 if( rc==SQLITE_OK && pSeg ){
2129 memset(pCsr, 0, sizeof(Fts3SegReaderCursor)); 2773 rc = fts3SegReaderCursorAppend(pCsr, pSeg);
2130
2131 /* If iLevel is less than 0, include a seg-reader for the pending-terms. */
2132 assert( isScan==0 || fts3HashCount(&p->pendingTerms)==0 );
2133 if( iLevel<0 && isScan==0 ){
2134 rc = sqlite3Fts3SegReaderPending(p, zTerm, nTerm, isPrefix, &pPending);
2135 if( rc==SQLITE_OK && pPending ){
2136 int nByte = (sizeof(Fts3SegReader *) * 16);
2137 pCsr->apSegment = (Fts3SegReader **)sqlite3_malloc(nByte);
2138 if( pCsr->apSegment==0 ){
2139 rc = SQLITE_NOMEM;
2140 }else{
2141 pCsr->apSegment[0] = pPending;
2142 pCsr->nSegment = 1;
2143 pPending = 0;
2144 }
2145 } 2774 }
2146 } 2775 }
2147 2776
2148 if( iLevel!=FTS3_SEGCURSOR_PENDING ){ 2777 if( iLevel!=FTS3_SEGCURSOR_PENDING ){
2149 if( rc==SQLITE_OK ){ 2778 if( rc==SQLITE_OK ){
2150 rc = sqlite3Fts3AllSegdirs(p, iLevel, &pStmt); 2779 rc = sqlite3Fts3AllSegdirs(p, iLangid, iIndex, iLevel, &pStmt);
2151 } 2780 }
2781
2152 while( rc==SQLITE_OK && SQLITE_ROW==(rc = sqlite3_step(pStmt)) ){ 2782 while( rc==SQLITE_OK && SQLITE_ROW==(rc = sqlite3_step(pStmt)) ){
2783 Fts3SegReader *pSeg = 0;
2153 2784
2154 /* Read the values returned by the SELECT into local variables. */ 2785 /* Read the values returned by the SELECT into local variables. */
2155 sqlite3_int64 iStartBlock = sqlite3_column_int64(pStmt, 1); 2786 sqlite3_int64 iStartBlock = sqlite3_column_int64(pStmt, 1);
2156 sqlite3_int64 iLeavesEndBlock = sqlite3_column_int64(pStmt, 2); 2787 sqlite3_int64 iLeavesEndBlock = sqlite3_column_int64(pStmt, 2);
2157 sqlite3_int64 iEndBlock = sqlite3_column_int64(pStmt, 3); 2788 sqlite3_int64 iEndBlock = sqlite3_column_int64(pStmt, 3);
2158 int nRoot = sqlite3_column_bytes(pStmt, 4); 2789 int nRoot = sqlite3_column_bytes(pStmt, 4);
2159 char const *zRoot = sqlite3_column_blob(pStmt, 4); 2790 char const *zRoot = sqlite3_column_blob(pStmt, 4);
2160 2791
2161 /* If nSegment is a multiple of 16 the array needs to be extended. */
2162 if( (pCsr->nSegment%16)==0 ){
2163 Fts3SegReader **apNew;
2164 int nByte = (pCsr->nSegment + 16)*sizeof(Fts3SegReader*);
2165 apNew = (Fts3SegReader **)sqlite3_realloc(pCsr->apSegment, nByte);
2166 if( !apNew ){
2167 rc = SQLITE_NOMEM;
2168 goto finished;
2169 }
2170 pCsr->apSegment = apNew;
2171 }
2172
2173 /* If zTerm is not NULL, and this segment is not stored entirely on its 2792 /* If zTerm is not NULL, and this segment is not stored entirely on its
2174 ** root node, the range of leaves scanned can be reduced. Do this. */ 2793 ** root node, the range of leaves scanned can be reduced. Do this. */
2175 if( iStartBlock && zTerm ){ 2794 if( iStartBlock && zTerm ){
2176 sqlite3_int64 *pi = (isPrefix ? &iLeavesEndBlock : 0); 2795 sqlite3_int64 *pi = (isPrefix ? &iLeavesEndBlock : 0);
2177 rc = fts3SelectLeaf(p, zTerm, nTerm, zRoot, nRoot, &iStartBlock, pi); 2796 rc = fts3SelectLeaf(p, zTerm, nTerm, zRoot, nRoot, &iStartBlock, pi);
2178 if( rc!=SQLITE_OK ) goto finished; 2797 if( rc!=SQLITE_OK ) goto finished;
2179 if( isPrefix==0 && isScan==0 ) iLeavesEndBlock = iStartBlock; 2798 if( isPrefix==0 && isScan==0 ) iLeavesEndBlock = iStartBlock;
2180 } 2799 }
2181 2800
2182 rc = sqlite3Fts3SegReaderNew(iAge, iStartBlock, iLeavesEndBlock, 2801 rc = sqlite3Fts3SegReaderNew(pCsr->nSegment+1,
2183 iEndBlock, zRoot, nRoot, &pCsr->apSegment[pCsr->nSegment] 2802 (isPrefix==0 && isScan==0),
2803 iStartBlock, iLeavesEndBlock,
2804 iEndBlock, zRoot, nRoot, &pSeg
2184 ); 2805 );
2185 if( rc!=SQLITE_OK ) goto finished; 2806 if( rc!=SQLITE_OK ) goto finished;
2186 pCsr->nSegment++; 2807 rc = fts3SegReaderCursorAppend(pCsr, pSeg);
2187 iAge++;
2188 } 2808 }
2189 } 2809 }
2190 2810
2191 finished: 2811 finished:
2192 rc2 = sqlite3_reset(pStmt); 2812 rc2 = sqlite3_reset(pStmt);
2193 if( rc==SQLITE_DONE ) rc = rc2; 2813 if( rc==SQLITE_DONE ) rc = rc2;
2194 sqlite3Fts3SegReaderFree(pPending);
2195 2814
2196 return rc; 2815 return rc;
2197 } 2816 }
2198 2817
2818 /*
2819 ** Set up a cursor object for iterating through a full-text index or a
2820 ** single level therein.
2821 */
2822 int sqlite3Fts3SegReaderCursor(
2823 Fts3Table *p, /* FTS3 table handle */
2824 int iLangid, /* Language-id to search */
2825 int iIndex, /* Index to search (from 0 to p->nIndex-1) */
2826 int iLevel, /* Level of segments to scan */
2827 const char *zTerm, /* Term to query for */
2828 int nTerm, /* Size of zTerm in bytes */
2829 int isPrefix, /* True for a prefix search */
2830 int isScan, /* True to scan from zTerm to EOF */
2831 Fts3MultiSegReader *pCsr /* Cursor object to populate */
2832 ){
2833 assert( iIndex>=0 && iIndex<p->nIndex );
2834 assert( iLevel==FTS3_SEGCURSOR_ALL
2835 || iLevel==FTS3_SEGCURSOR_PENDING
2836 || iLevel>=0
2837 );
2838 assert( iLevel<FTS3_SEGDIR_MAXLEVEL );
2839 assert( FTS3_SEGCURSOR_ALL<0 && FTS3_SEGCURSOR_PENDING<0 );
2840 assert( isPrefix==0 || isScan==0 );
2199 2841
2842 memset(pCsr, 0, sizeof(Fts3MultiSegReader));
2843 return fts3SegReaderCursor(
2844 p, iLangid, iIndex, iLevel, zTerm, nTerm, isPrefix, isScan, pCsr
2845 );
2846 }
2847
2848 /*
2849 ** In addition to its current configuration, have the Fts3MultiSegReader
2850 ** passed as the 4th argument also scan the doclist for term zTerm/nTerm.
2851 **
2852 ** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
2853 */
2854 static int fts3SegReaderCursorAddZero(
2855 Fts3Table *p, /* FTS virtual table handle */
2856 int iLangid,
2857 const char *zTerm, /* Term to scan doclist of */
2858 int nTerm, /* Number of bytes in zTerm */
2859 Fts3MultiSegReader *pCsr /* Fts3MultiSegReader to modify */
2860 ){
2861 return fts3SegReaderCursor(p,
2862 iLangid, 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0,pCsr
2863 );
2864 }
2865
2866 /*
2867 ** Open an Fts3MultiSegReader to scan the doclist for term zTerm/nTerm. Or,
2868 ** if isPrefix is true, to scan the doclist for all terms for which
2869 ** zTerm/nTerm is a prefix. If successful, return SQLITE_OK and write
2870 ** a pointer to the new Fts3MultiSegReader to *ppSegcsr. Otherwise, return
2871 ** an SQLite error code.
2872 **
2873 ** It is the responsibility of the caller to free this object by eventually
2874 ** passing it to fts3SegReaderCursorFree()
2875 **
2876 ** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
2877 ** Output parameter *ppSegcsr is set to 0 if an error occurs.
2878 */
2200 static int fts3TermSegReaderCursor( 2879 static int fts3TermSegReaderCursor(
2201 Fts3Cursor *pCsr, /* Virtual table cursor handle */ 2880 Fts3Cursor *pCsr, /* Virtual table cursor handle */
2202 const char *zTerm, /* Term to query for */ 2881 const char *zTerm, /* Term to query for */
2203 int nTerm, /* Size of zTerm in bytes */ 2882 int nTerm, /* Size of zTerm in bytes */
2204 int isPrefix, /* True for a prefix search */ 2883 int isPrefix, /* True for a prefix search */
2205 Fts3SegReaderCursor **ppSegcsr /* OUT: Allocated seg-reader cursor */ 2884 Fts3MultiSegReader **ppSegcsr /* OUT: Allocated seg-reader cursor */
2206 ){ 2885 ){
2207 Fts3SegReaderCursor *pSegcsr; /* Object to allocate and return */ 2886 Fts3MultiSegReader *pSegcsr; /* Object to allocate and return */
2208 int rc = SQLITE_NOMEM; /* Return code */ 2887 int rc = SQLITE_NOMEM; /* Return code */
2209 2888
2210 pSegcsr = sqlite3_malloc(sizeof(Fts3SegReaderCursor)); 2889 pSegcsr = sqlite3_malloc(sizeof(Fts3MultiSegReader));
2211 if( pSegcsr ){ 2890 if( pSegcsr ){
2891 int i;
2892 int bFound = 0; /* True once an index has been found */
2212 Fts3Table *p = (Fts3Table *)pCsr->base.pVtab; 2893 Fts3Table *p = (Fts3Table *)pCsr->base.pVtab;
2213 int i; 2894
2214 int nCost = 0; 2895 if( isPrefix ){
2215 rc = sqlite3Fts3SegReaderCursor( 2896 for(i=1; bFound==0 && i<p->nIndex; i++){
2216 p, FTS3_SEGCURSOR_ALL, zTerm, nTerm, isPrefix, 0, pSegcsr); 2897 if( p->aIndex[i].nPrefix==nTerm ){
2217 2898 bFound = 1;
2218 for(i=0; rc==SQLITE_OK && i<pSegcsr->nSegment; i++){ 2899 rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
2219 rc = sqlite3Fts3SegReaderCost(pCsr, pSegcsr->apSegment[i], &nCost); 2900 i, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 0, 0, pSegcsr
2901 );
2902 pSegcsr->bLookup = 1;
2903 }
2904 }
2905
2906 for(i=1; bFound==0 && i<p->nIndex; i++){
2907 if( p->aIndex[i].nPrefix==nTerm+1 ){
2908 bFound = 1;
2909 rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
2910 i, FTS3_SEGCURSOR_ALL, zTerm, nTerm, 1, 0, pSegcsr
2911 );
2912 if( rc==SQLITE_OK ){
2913 rc = fts3SegReaderCursorAddZero(
2914 p, pCsr->iLangid, zTerm, nTerm, pSegcsr
2915 );
2916 }
2917 }
2918 }
2220 } 2919 }
2221 pSegcsr->nCost = nCost; 2920
2921 if( bFound==0 ){
2922 rc = sqlite3Fts3SegReaderCursor(p, pCsr->iLangid,
2923 0, FTS3_SEGCURSOR_ALL, zTerm, nTerm, isPrefix, 0, pSegcsr
2924 );
2925 pSegcsr->bLookup = !isPrefix;
2926 }
2222 } 2927 }
2223 2928
2224 *ppSegcsr = pSegcsr; 2929 *ppSegcsr = pSegcsr;
2225 return rc; 2930 return rc;
2226 } 2931 }
2227 2932
2228 static void fts3SegReaderCursorFree(Fts3SegReaderCursor *pSegcsr){ 2933 /*
2934 ** Free an Fts3MultiSegReader allocated by fts3TermSegReaderCursor().
2935 */
2936 static void fts3SegReaderCursorFree(Fts3MultiSegReader *pSegcsr){
2229 sqlite3Fts3SegReaderFinish(pSegcsr); 2937 sqlite3Fts3SegReaderFinish(pSegcsr);
2230 sqlite3_free(pSegcsr); 2938 sqlite3_free(pSegcsr);
2231 } 2939 }
2232 2940
2233 /* 2941 /*
2234 ** This function retreives the doclist for the specified term (or term 2942 ** This function retrieves the doclist for the specified term (or term
2235 ** prefix) from the database. 2943 ** prefix) from the database.
2236 **
2237 ** The returned doclist may be in one of two formats, depending on the
2238 ** value of parameter isReqPos. If isReqPos is zero, then the doclist is
2239 ** a sorted list of delta-compressed docids (a bare doclist). If isReqPos
2240 ** is non-zero, then the returned list is in the same format as is stored
2241 ** in the database without the found length specifier at the start of on-disk
2242 ** doclists.
2243 */ 2944 */
2244 static int fts3TermSelect( 2945 static int fts3TermSelect(
2245 Fts3Table *p, /* Virtual table handle */ 2946 Fts3Table *p, /* Virtual table handle */
2246 Fts3PhraseToken *pTok, /* Token to query for */ 2947 Fts3PhraseToken *pTok, /* Token to query for */
2247 int iColumn, /* Column to query (or -ve for all columns) */ 2948 int iColumn, /* Column to query (or -ve for all columns) */
2248 int isReqPos, /* True to include position lists in output */
2249 int *pnOut, /* OUT: Size of buffer at *ppOut */ 2949 int *pnOut, /* OUT: Size of buffer at *ppOut */
2250 char **ppOut /* OUT: Malloced result buffer */ 2950 char **ppOut /* OUT: Malloced result buffer */
2251 ){ 2951 ){
2252 int rc; /* Return code */ 2952 int rc; /* Return code */
2253 Fts3SegReaderCursor *pSegcsr; /* Seg-reader cursor for this term */ 2953 Fts3MultiSegReader *pSegcsr; /* Seg-reader cursor for this term */
2254 TermSelect tsc; /* Context object for fts3TermSelectCb() */ 2954 TermSelect tsc; /* Object for pair-wise doclist merging */
2255 Fts3SegFilter filter; /* Segment term filter configuration */ 2955 Fts3SegFilter filter; /* Segment term filter configuration */
2256 2956
2257 pSegcsr = pTok->pSegcsr; 2957 pSegcsr = pTok->pSegcsr;
2258 memset(&tsc, 0, sizeof(TermSelect)); 2958 memset(&tsc, 0, sizeof(TermSelect));
2259 tsc.isReqPos = isReqPos;
2260 2959
2261 filter.flags = FTS3_SEGMENT_IGNORE_EMPTY 2960 filter.flags = FTS3_SEGMENT_IGNORE_EMPTY | FTS3_SEGMENT_REQUIRE_POS
2262 | (pTok->isPrefix ? FTS3_SEGMENT_PREFIX : 0) 2961 | (pTok->isPrefix ? FTS3_SEGMENT_PREFIX : 0)
2263 | (isReqPos ? FTS3_SEGMENT_REQUIRE_POS : 0) 2962 | (pTok->bFirst ? FTS3_SEGMENT_FIRST : 0)
2264 | (iColumn<p->nColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0); 2963 | (iColumn<p->nColumn ? FTS3_SEGMENT_COLUMN_FILTER : 0);
2265 filter.iCol = iColumn; 2964 filter.iCol = iColumn;
2266 filter.zTerm = pTok->z; 2965 filter.zTerm = pTok->z;
2267 filter.nTerm = pTok->n; 2966 filter.nTerm = pTok->n;
2268 2967
2269 rc = sqlite3Fts3SegReaderStart(p, pSegcsr, &filter); 2968 rc = sqlite3Fts3SegReaderStart(p, pSegcsr, &filter);
2270 while( SQLITE_OK==rc 2969 while( SQLITE_OK==rc
2271 && SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pSegcsr)) 2970 && SQLITE_ROW==(rc = sqlite3Fts3SegReaderStep(p, pSegcsr))
2272 ){ 2971 ){
2273 rc = fts3TermSelectCb(p, (void *)&tsc, 2972 rc = fts3TermSelectMerge(p, &tsc, pSegcsr->aDoclist, pSegcsr->nDoclist);
2274 pSegcsr->zTerm, pSegcsr->nTerm, pSegcsr->aDoclist, pSegcsr->nDoclist
2275 );
2276 } 2973 }
2277 2974
2278 if( rc==SQLITE_OK ){ 2975 if( rc==SQLITE_OK ){
2279 rc = fts3TermSelectMerge(&tsc); 2976 rc = fts3TermSelectFinishMerge(p, &tsc);
2280 } 2977 }
2281 if( rc==SQLITE_OK ){ 2978 if( rc==SQLITE_OK ){
2282 *ppOut = tsc.aaOutput[0]; 2979 *ppOut = tsc.aaOutput[0];
2283 *pnOut = tsc.anOutput[0]; 2980 *pnOut = tsc.anOutput[0];
2284 }else{ 2981 }else{
2285 int i; 2982 int i;
2286 for(i=0; i<SizeofArray(tsc.aaOutput); i++){ 2983 for(i=0; i<SizeofArray(tsc.aaOutput); i++){
2287 sqlite3_free(tsc.aaOutput[i]); 2984 sqlite3_free(tsc.aaOutput[i]);
2288 } 2985 }
2289 } 2986 }
2290 2987
2291 fts3SegReaderCursorFree(pSegcsr); 2988 fts3SegReaderCursorFree(pSegcsr);
2292 pTok->pSegcsr = 0; 2989 pTok->pSegcsr = 0;
2293 return rc; 2990 return rc;
2294 } 2991 }
2295 2992
2296 /* 2993 /*
2297 ** This function counts the total number of docids in the doclist stored 2994 ** This function counts the total number of docids in the doclist stored
2298 ** in buffer aList[], size nList bytes. 2995 ** in buffer aList[], size nList bytes.
2299 ** 2996 **
2300 ** If the isPoslist argument is true, then it is assumed that the doclist 2997 ** If the isPoslist argument is true, then it is assumed that the doclist
2301 ** contains a position-list following each docid. Otherwise, it is assumed 2998 ** contains a position-list following each docid. Otherwise, it is assumed
2302 ** that the doclist is simply a list of docids stored as delta encoded 2999 ** that the doclist is simply a list of docids stored as delta encoded
2303 ** varints. 3000 ** varints.
2304 */ 3001 */
2305 static int fts3DoclistCountDocids(int isPoslist, char *aList, int nList){ 3002 static int fts3DoclistCountDocids(char *aList, int nList){
2306 int nDoc = 0; /* Return value */ 3003 int nDoc = 0; /* Return value */
2307 if( aList ){ 3004 if( aList ){
2308 char *aEnd = &aList[nList]; /* Pointer to one byte after EOF */ 3005 char *aEnd = &aList[nList]; /* Pointer to one byte after EOF */
2309 char *p = aList; /* Cursor */ 3006 char *p = aList; /* Cursor */
2310 if( !isPoslist ){ 3007 while( p<aEnd ){
2311 /* The number of docids in the list is the same as the number of 3008 nDoc++;
2312 ** varints. In FTS3 a varint consists of a single byte with the 0x80 3009 while( (*p++)&0x80 ); /* Skip docid varint */
2313 ** bit cleared and zero or more bytes with the 0x80 bit set. So to 3010 fts3PoslistCopy(0, &p); /* Skip over position list */
2314 ** count the varints in the buffer, just count the number of bytes
2315 ** with the 0x80 bit clear. */
2316 while( p<aEnd ) nDoc += (((*p++)&0x80)==0);
2317 }else{
2318 while( p<aEnd ){
2319 nDoc++;
2320 while( (*p++)&0x80 ); /* Skip docid varint */
2321 fts3PoslistCopy(0, &p); /* Skip over position list */
2322 }
2323 } 3011 }
2324 } 3012 }
2325 3013
2326 return nDoc; 3014 return nDoc;
2327 } 3015 }
2328 3016
2329 /* 3017 /*
2330 ** Call sqlite3Fts3DeferToken() for each token in the expression pExpr.
2331 */
2332 static int fts3DeferExpression(Fts3Cursor *pCsr, Fts3Expr *pExpr){
2333 int rc = SQLITE_OK;
2334 if( pExpr ){
2335 rc = fts3DeferExpression(pCsr, pExpr->pLeft);
2336 if( rc==SQLITE_OK ){
2337 rc = fts3DeferExpression(pCsr, pExpr->pRight);
2338 }
2339 if( pExpr->eType==FTSQUERY_PHRASE ){
2340 int iCol = pExpr->pPhrase->iColumn;
2341 int i;
2342 for(i=0; rc==SQLITE_OK && i<pExpr->pPhrase->nToken; i++){
2343 Fts3PhraseToken *pToken = &pExpr->pPhrase->aToken[i];
2344 if( pToken->pDeferred==0 ){
2345 rc = sqlite3Fts3DeferToken(pCsr, pToken, iCol);
2346 }
2347 }
2348 }
2349 }
2350 return rc;
2351 }
2352
2353 /*
2354 ** This function removes the position information from a doclist. When
2355 ** called, buffer aList (size *pnList bytes) contains a doclist that includes
2356 ** position information. This function removes the position information so
2357 ** that aList contains only docids, and adjusts *pnList to reflect the new
2358 ** (possibly reduced) size of the doclist.
2359 */
2360 static void fts3DoclistStripPositions(
2361 char *aList, /* IN/OUT: Buffer containing doclist */
2362 int *pnList /* IN/OUT: Size of doclist in bytes */
2363 ){
2364 if( aList ){
2365 char *aEnd = &aList[*pnList]; /* Pointer to one byte after EOF */
2366 char *p = aList; /* Input cursor */
2367 char *pOut = aList; /* Output cursor */
2368
2369 while( p<aEnd ){
2370 sqlite3_int64 delta;
2371 p += sqlite3Fts3GetVarint(p, &delta);
2372 fts3PoslistCopy(0, &p);
2373 pOut += sqlite3Fts3PutVarint(pOut, delta);
2374 }
2375
2376 *pnList = (int)(pOut - aList);
2377 }
2378 }
2379
2380 /*
2381 ** Return a DocList corresponding to the phrase *pPhrase.
2382 **
2383 ** If this function returns SQLITE_OK, but *pnOut is set to a negative value,
2384 ** then no tokens in the phrase were looked up in the full-text index. This
2385 ** is only possible when this function is called from within xFilter(). The
2386 ** caller should assume that all documents match the phrase. The actual
2387 ** filtering will take place in xNext().
2388 */
2389 static int fts3PhraseSelect(
2390 Fts3Cursor *pCsr, /* Virtual table cursor handle */
2391 Fts3Phrase *pPhrase, /* Phrase to return a doclist for */
2392 int isReqPos, /* True if output should contain positions */
2393 char **paOut, /* OUT: Pointer to malloc'd result buffer */
2394 int *pnOut /* OUT: Size of buffer at *paOut */
2395 ){
2396 char *pOut = 0;
2397 int nOut = 0;
2398 int rc = SQLITE_OK;
2399 int ii;
2400 int iCol = pPhrase->iColumn;
2401 int isTermPos = (pPhrase->nToken>1 || isReqPos);
2402 Fts3Table *p = (Fts3Table *)pCsr->base.pVtab;
2403 int isFirst = 1;
2404
2405 int iPrevTok = 0;
2406 int nDoc = 0;
2407
2408 /* If this is an xFilter() evaluation, create a segment-reader for each
2409 ** phrase token. Or, if this is an xNext() or snippet/offsets/matchinfo
2410 ** evaluation, only create segment-readers if there are no Fts3DeferredToken
2411 ** objects attached to the phrase-tokens.
2412 */
2413 for(ii=0; ii<pPhrase->nToken; ii++){
2414 Fts3PhraseToken *pTok = &pPhrase->aToken[ii];
2415 if( pTok->pSegcsr==0 ){
2416 if( (pCsr->eEvalmode==FTS3_EVAL_FILTER)
2417 || (pCsr->eEvalmode==FTS3_EVAL_NEXT && pCsr->pDeferred==0)
2418 || (pCsr->eEvalmode==FTS3_EVAL_MATCHINFO && pTok->bFulltext)
2419 ){
2420 rc = fts3TermSegReaderCursor(
2421 pCsr, pTok->z, pTok->n, pTok->isPrefix, &pTok->pSegcsr
2422 );
2423 if( rc!=SQLITE_OK ) return rc;
2424 }
2425 }
2426 }
2427
2428 for(ii=0; ii<pPhrase->nToken; ii++){
2429 Fts3PhraseToken *pTok; /* Token to find doclist for */
2430 int iTok = 0; /* The token being queried this iteration */
2431 char *pList = 0; /* Pointer to token doclist */
2432 int nList = 0; /* Size of buffer at pList */
2433
2434 /* Select a token to process. If this is an xFilter() call, then tokens
2435 ** are processed in order from least to most costly. Otherwise, tokens
2436 ** are processed in the order in which they occur in the phrase.
2437 */
2438 if( pCsr->eEvalmode==FTS3_EVAL_MATCHINFO ){
2439 assert( isReqPos );
2440 iTok = ii;
2441 pTok = &pPhrase->aToken[iTok];
2442 if( pTok->bFulltext==0 ) continue;
2443 }else if( pCsr->eEvalmode==FTS3_EVAL_NEXT || isReqPos ){
2444 iTok = ii;
2445 pTok = &pPhrase->aToken[iTok];
2446 }else{
2447 int nMinCost = 0x7FFFFFFF;
2448 int jj;
2449
2450 /* Find the remaining token with the lowest cost. */
2451 for(jj=0; jj<pPhrase->nToken; jj++){
2452 Fts3SegReaderCursor *pSegcsr = pPhrase->aToken[jj].pSegcsr;
2453 if( pSegcsr && pSegcsr->nCost<nMinCost ){
2454 iTok = jj;
2455 nMinCost = pSegcsr->nCost;
2456 }
2457 }
2458 pTok = &pPhrase->aToken[iTok];
2459
2460 /* This branch is taken if it is determined that loading the doclist
2461 ** for the next token would require more IO than loading all documents
2462 ** currently identified by doclist pOut/nOut. No further doclists will
2463 ** be loaded from the full-text index for this phrase.
2464 */
2465 if( nMinCost>nDoc && ii>0 ){
2466 rc = fts3DeferExpression(pCsr, pCsr->pExpr);
2467 break;
2468 }
2469 }
2470
2471 if( pCsr->eEvalmode==FTS3_EVAL_NEXT && pTok->pDeferred ){
2472 rc = fts3DeferredTermSelect(pTok->pDeferred, isTermPos, &nList, &pList);
2473 }else{
2474 if( pTok->pSegcsr ){
2475 rc = fts3TermSelect(p, pTok, iCol, isTermPos, &nList, &pList);
2476 }
2477 pTok->bFulltext = 1;
2478 }
2479 assert( rc!=SQLITE_OK || pCsr->eEvalmode || pTok->pSegcsr==0 );
2480 if( rc!=SQLITE_OK ) break;
2481
2482 if( isFirst ){
2483 pOut = pList;
2484 nOut = nList;
2485 if( pCsr->eEvalmode==FTS3_EVAL_FILTER && pPhrase->nToken>1 ){
2486 nDoc = fts3DoclistCountDocids(1, pOut, nOut);
2487 }
2488 isFirst = 0;
2489 iPrevTok = iTok;
2490 }else{
2491 /* Merge the new term list and the current output. */
2492 char *aLeft, *aRight;
2493 int nLeft, nRight;
2494 int nDist;
2495 int mt;
2496
2497 /* If this is the final token of the phrase, and positions were not
2498 ** requested by the caller, use MERGE_PHRASE instead of POS_PHRASE.
2499 ** This drops the position information from the output list.
2500 */
2501 mt = MERGE_POS_PHRASE;
2502 if( ii==pPhrase->nToken-1 && !isReqPos ) mt = MERGE_PHRASE;
2503
2504 assert( iPrevTok!=iTok );
2505 if( iPrevTok<iTok ){
2506 aLeft = pOut;
2507 nLeft = nOut;
2508 aRight = pList;
2509 nRight = nList;
2510 nDist = iTok-iPrevTok;
2511 iPrevTok = iTok;
2512 }else{
2513 aRight = pOut;
2514 nRight = nOut;
2515 aLeft = pList;
2516 nLeft = nList;
2517 nDist = iPrevTok-iTok;
2518 }
2519 pOut = aRight;
2520 fts3DoclistMerge(
2521 mt, nDist, 0, pOut, &nOut, aLeft, nLeft, aRight, nRight, &nDoc
2522 );
2523 sqlite3_free(aLeft);
2524 }
2525 assert( nOut==0 || pOut!=0 );
2526 }
2527
2528 if( rc==SQLITE_OK ){
2529 if( ii!=pPhrase->nToken ){
2530 assert( pCsr->eEvalmode==FTS3_EVAL_FILTER && isReqPos==0 );
2531 fts3DoclistStripPositions(pOut, &nOut);
2532 }
2533 *paOut = pOut;
2534 *pnOut = nOut;
2535 }else{
2536 sqlite3_free(pOut);
2537 }
2538 return rc;
2539 }
2540
2541 /*
2542 ** This function merges two doclists according to the requirements of a
2543 ** NEAR operator.
2544 **
2545 ** Both input doclists must include position information. The output doclist
2546 ** includes position information if the first argument to this function
2547 ** is MERGE_POS_NEAR, or does not if it is MERGE_NEAR.
2548 */
2549 static int fts3NearMerge(
2550 int mergetype, /* MERGE_POS_NEAR or MERGE_NEAR */
2551 int nNear, /* Parameter to NEAR operator */
2552 int nTokenLeft, /* Number of tokens in LHS phrase arg */
2553 char *aLeft, /* Doclist for LHS (incl. positions) */
2554 int nLeft, /* Size of LHS doclist in bytes */
2555 int nTokenRight, /* As nTokenLeft */
2556 char *aRight, /* As aLeft */
2557 int nRight, /* As nRight */
2558 char **paOut, /* OUT: Results of merge (malloced) */
2559 int *pnOut /* OUT: Sized of output buffer */
2560 ){
2561 char *aOut; /* Buffer to write output doclist to */
2562 int rc; /* Return code */
2563
2564 assert( mergetype==MERGE_POS_NEAR || MERGE_NEAR );
2565
2566 aOut = sqlite3_malloc(nLeft+nRight+1);
2567 if( aOut==0 ){
2568 rc = SQLITE_NOMEM;
2569 }else{
2570 rc = fts3DoclistMerge(mergetype, nNear+nTokenRight, nNear+nTokenLeft,
2571 aOut, pnOut, aLeft, nLeft, aRight, nRight, 0
2572 );
2573 if( rc!=SQLITE_OK ){
2574 sqlite3_free(aOut);
2575 aOut = 0;
2576 }
2577 }
2578
2579 *paOut = aOut;
2580 return rc;
2581 }
2582
2583 /*
2584 ** This function is used as part of the processing for the snippet() and
2585 ** offsets() functions.
2586 **
2587 ** Both pLeft and pRight are expression nodes of type FTSQUERY_PHRASE. Both
2588 ** have their respective doclists (including position information) loaded
2589 ** in Fts3Expr.aDoclist/nDoclist. This function removes all entries from
2590 ** each doclist that are not within nNear tokens of a corresponding entry
2591 ** in the other doclist.
2592 */
2593 int sqlite3Fts3ExprNearTrim(Fts3Expr *pLeft, Fts3Expr *pRight, int nNear){
2594 int rc; /* Return code */
2595
2596 assert( pLeft->eType==FTSQUERY_PHRASE );
2597 assert( pRight->eType==FTSQUERY_PHRASE );
2598 assert( pLeft->isLoaded && pRight->isLoaded );
2599
2600 if( pLeft->aDoclist==0 || pRight->aDoclist==0 ){
2601 sqlite3_free(pLeft->aDoclist);
2602 sqlite3_free(pRight->aDoclist);
2603 pRight->aDoclist = 0;
2604 pLeft->aDoclist = 0;
2605 rc = SQLITE_OK;
2606 }else{
2607 char *aOut; /* Buffer in which to assemble new doclist */
2608 int nOut; /* Size of buffer aOut in bytes */
2609
2610 rc = fts3NearMerge(MERGE_POS_NEAR, nNear,
2611 pLeft->pPhrase->nToken, pLeft->aDoclist, pLeft->nDoclist,
2612 pRight->pPhrase->nToken, pRight->aDoclist, pRight->nDoclist,
2613 &aOut, &nOut
2614 );
2615 if( rc!=SQLITE_OK ) return rc;
2616 sqlite3_free(pRight->aDoclist);
2617 pRight->aDoclist = aOut;
2618 pRight->nDoclist = nOut;
2619
2620 rc = fts3NearMerge(MERGE_POS_NEAR, nNear,
2621 pRight->pPhrase->nToken, pRight->aDoclist, pRight->nDoclist,
2622 pLeft->pPhrase->nToken, pLeft->aDoclist, pLeft->nDoclist,
2623 &aOut, &nOut
2624 );
2625 sqlite3_free(pLeft->aDoclist);
2626 pLeft->aDoclist = aOut;
2627 pLeft->nDoclist = nOut;
2628 }
2629 return rc;
2630 }
2631
2632
2633 /*
2634 ** Allocate an Fts3SegReaderArray for each token in the expression pExpr.
2635 ** The allocated objects are stored in the Fts3PhraseToken.pArray member
2636 ** variables of each token structure.
2637 */
2638 static int fts3ExprAllocateSegReaders(
2639 Fts3Cursor *pCsr, /* FTS3 table */
2640 Fts3Expr *pExpr, /* Expression to create seg-readers for */
2641 int *pnExpr /* OUT: Number of AND'd expressions */
2642 ){
2643 int rc = SQLITE_OK; /* Return code */
2644
2645 assert( pCsr->eEvalmode==FTS3_EVAL_FILTER );
2646 if( pnExpr && pExpr->eType!=FTSQUERY_AND ){
2647 (*pnExpr)++;
2648 pnExpr = 0;
2649 }
2650
2651 if( pExpr->eType==FTSQUERY_PHRASE ){
2652 Fts3Phrase *pPhrase = pExpr->pPhrase;
2653 int ii;
2654
2655 for(ii=0; rc==SQLITE_OK && ii<pPhrase->nToken; ii++){
2656 Fts3PhraseToken *pTok = &pPhrase->aToken[ii];
2657 if( pTok->pSegcsr==0 ){
2658 rc = fts3TermSegReaderCursor(
2659 pCsr, pTok->z, pTok->n, pTok->isPrefix, &pTok->pSegcsr
2660 );
2661 }
2662 }
2663 }else{
2664 rc = fts3ExprAllocateSegReaders(pCsr, pExpr->pLeft, pnExpr);
2665 if( rc==SQLITE_OK ){
2666 rc = fts3ExprAllocateSegReaders(pCsr, pExpr->pRight, pnExpr);
2667 }
2668 }
2669 return rc;
2670 }
2671
2672 /*
2673 ** Free the Fts3SegReaderArray objects associated with each token in the
2674 ** expression pExpr. In other words, this function frees the resources
2675 ** allocated by fts3ExprAllocateSegReaders().
2676 */
2677 static void fts3ExprFreeSegReaders(Fts3Expr *pExpr){
2678 if( pExpr ){
2679 Fts3Phrase *pPhrase = pExpr->pPhrase;
2680 if( pPhrase ){
2681 int kk;
2682 for(kk=0; kk<pPhrase->nToken; kk++){
2683 fts3SegReaderCursorFree(pPhrase->aToken[kk].pSegcsr);
2684 pPhrase->aToken[kk].pSegcsr = 0;
2685 }
2686 }
2687 fts3ExprFreeSegReaders(pExpr->pLeft);
2688 fts3ExprFreeSegReaders(pExpr->pRight);
2689 }
2690 }
2691
2692 /*
2693 ** Return the sum of the costs of all tokens in the expression pExpr. This
2694 ** function must be called after Fts3SegReaderArrays have been allocated
2695 ** for all tokens using fts3ExprAllocateSegReaders().
2696 */
2697 static int fts3ExprCost(Fts3Expr *pExpr){
2698 int nCost; /* Return value */
2699 if( pExpr->eType==FTSQUERY_PHRASE ){
2700 Fts3Phrase *pPhrase = pExpr->pPhrase;
2701 int ii;
2702 nCost = 0;
2703 for(ii=0; ii<pPhrase->nToken; ii++){
2704 Fts3SegReaderCursor *pSegcsr = pPhrase->aToken[ii].pSegcsr;
2705 if( pSegcsr ) nCost += pSegcsr->nCost;
2706 }
2707 }else{
2708 nCost = fts3ExprCost(pExpr->pLeft) + fts3ExprCost(pExpr->pRight);
2709 }
2710 return nCost;
2711 }
2712
2713 /*
2714 ** The following is a helper function (and type) for fts3EvalExpr(). It
2715 ** must be called after Fts3SegReaders have been allocated for every token
2716 ** in the expression. See the context it is called from in fts3EvalExpr()
2717 ** for further explanation.
2718 */
2719 typedef struct ExprAndCost ExprAndCost;
2720 struct ExprAndCost {
2721 Fts3Expr *pExpr;
2722 int nCost;
2723 };
2724 static void fts3ExprAssignCosts(
2725 Fts3Expr *pExpr, /* Expression to create seg-readers for */
2726 ExprAndCost **ppExprCost /* OUT: Write to *ppExprCost */
2727 ){
2728 if( pExpr->eType==FTSQUERY_AND ){
2729 fts3ExprAssignCosts(pExpr->pLeft, ppExprCost);
2730 fts3ExprAssignCosts(pExpr->pRight, ppExprCost);
2731 }else{
2732 (*ppExprCost)->pExpr = pExpr;
2733 (*ppExprCost)->nCost = fts3ExprCost(pExpr);
2734 (*ppExprCost)++;
2735 }
2736 }
2737
2738 /*
2739 ** Evaluate the full-text expression pExpr against FTS3 table pTab. Store
2740 ** the resulting doclist in *paOut and *pnOut. This routine mallocs for
2741 ** the space needed to store the output. The caller is responsible for
2742 ** freeing the space when it has finished.
2743 **
2744 ** This function is called in two distinct contexts:
2745 **
2746 ** * From within the virtual table xFilter() method. In this case, the
2747 ** output doclist contains entries for all rows in the table, based on
2748 ** data read from the full-text index.
2749 **
2750 ** In this case, if the query expression contains one or more tokens that
2751 ** are very common, then the returned doclist may contain a superset of
2752 ** the documents that actually match the expression.
2753 **
2754 ** * From within the virtual table xNext() method. This call is only made
2755 ** if the call from within xFilter() found that there were very common
2756 ** tokens in the query expression and did return a superset of the
2757 ** matching documents. In this case the returned doclist contains only
2758 ** entries that correspond to the current row of the table. Instead of
2759 ** reading the data for each token from the full-text index, the data is
2760 ** already available in-memory in the Fts3PhraseToken.pDeferred structures.
2761 ** See fts3EvalDeferred() for how it gets there.
2762 **
2763 ** In the first case above, Fts3Cursor.doDeferred==0. In the second (if it is
2764 ** required) Fts3Cursor.doDeferred==1.
2765 **
2766 ** If the SQLite invokes the snippet(), offsets() or matchinfo() function
2767 ** as part of a SELECT on an FTS3 table, this function is called on each
2768 ** individual phrase expression in the query. If there were very common tokens
2769 ** found in the xFilter() call, then this function is called once for phrase
2770 ** for each row visited, and the returned doclist contains entries for the
2771 ** current row only. Otherwise, if there were no very common tokens, then this
2772 ** function is called once only for each phrase in the query and the returned
2773 ** doclist contains entries for all rows of the table.
2774 **
2775 ** Fts3Cursor.doDeferred==1 when this function is called on phrases as a
2776 ** result of a snippet(), offsets() or matchinfo() invocation.
2777 */
2778 static int fts3EvalExpr(
2779 Fts3Cursor *p, /* Virtual table cursor handle */
2780 Fts3Expr *pExpr, /* Parsed fts3 expression */
2781 char **paOut, /* OUT: Pointer to malloc'd result buffer */
2782 int *pnOut, /* OUT: Size of buffer at *paOut */
2783 int isReqPos /* Require positions in output buffer */
2784 ){
2785 int rc = SQLITE_OK; /* Return code */
2786
2787 /* Zero the output parameters. */
2788 *paOut = 0;
2789 *pnOut = 0;
2790
2791 if( pExpr ){
2792 assert( pExpr->eType==FTSQUERY_NEAR || pExpr->eType==FTSQUERY_OR
2793 || pExpr->eType==FTSQUERY_AND || pExpr->eType==FTSQUERY_NOT
2794 || pExpr->eType==FTSQUERY_PHRASE
2795 );
2796 assert( pExpr->eType==FTSQUERY_PHRASE || isReqPos==0 );
2797
2798 if( pExpr->eType==FTSQUERY_PHRASE ){
2799 rc = fts3PhraseSelect(p, pExpr->pPhrase,
2800 isReqPos || (pExpr->pParent && pExpr->pParent->eType==FTSQUERY_NEAR),
2801 paOut, pnOut
2802 );
2803 fts3ExprFreeSegReaders(pExpr);
2804 }else if( p->eEvalmode==FTS3_EVAL_FILTER && pExpr->eType==FTSQUERY_AND ){
2805 ExprAndCost *aExpr = 0; /* Array of AND'd expressions and costs */
2806 int nExpr = 0; /* Size of aExpr[] */
2807 char *aRet = 0; /* Doclist to return to caller */
2808 int nRet = 0; /* Length of aRet[] in bytes */
2809 int nDoc = 0x7FFFFFFF;
2810
2811 assert( !isReqPos );
2812
2813 rc = fts3ExprAllocateSegReaders(p, pExpr, &nExpr);
2814 if( rc==SQLITE_OK ){
2815 assert( nExpr>1 );
2816 aExpr = sqlite3_malloc(sizeof(ExprAndCost) * nExpr);
2817 if( !aExpr ) rc = SQLITE_NOMEM;
2818 }
2819 if( rc==SQLITE_OK ){
2820 int ii; /* Used to iterate through expressions */
2821
2822 fts3ExprAssignCosts(pExpr, &aExpr);
2823 aExpr -= nExpr;
2824 for(ii=0; ii<nExpr; ii++){
2825 char *aNew;
2826 int nNew;
2827 int jj;
2828 ExprAndCost *pBest = 0;
2829
2830 for(jj=0; jj<nExpr; jj++){
2831 ExprAndCost *pCand = &aExpr[jj];
2832 if( pCand->pExpr && (pBest==0 || pCand->nCost<pBest->nCost) ){
2833 pBest = pCand;
2834 }
2835 }
2836
2837 if( pBest->nCost>nDoc ){
2838 rc = fts3DeferExpression(p, p->pExpr);
2839 break;
2840 }else{
2841 rc = fts3EvalExpr(p, pBest->pExpr, &aNew, &nNew, 0);
2842 if( rc!=SQLITE_OK ) break;
2843 pBest->pExpr = 0;
2844 if( ii==0 ){
2845 aRet = aNew;
2846 nRet = nNew;
2847 nDoc = fts3DoclistCountDocids(0, aRet, nRet);
2848 }else{
2849 fts3DoclistMerge(
2850 MERGE_AND, 0, 0, aRet, &nRet, aRet, nRet, aNew, nNew, &nDoc
2851 );
2852 sqlite3_free(aNew);
2853 }
2854 }
2855 }
2856 }
2857
2858 if( rc==SQLITE_OK ){
2859 *paOut = aRet;
2860 *pnOut = nRet;
2861 }else{
2862 assert( *paOut==0 );
2863 sqlite3_free(aRet);
2864 }
2865 sqlite3_free(aExpr);
2866 fts3ExprFreeSegReaders(pExpr);
2867
2868 }else{
2869 char *aLeft;
2870 char *aRight;
2871 int nLeft;
2872 int nRight;
2873
2874 assert( pExpr->eType==FTSQUERY_NEAR
2875 || pExpr->eType==FTSQUERY_OR
2876 || pExpr->eType==FTSQUERY_NOT
2877 || (pExpr->eType==FTSQUERY_AND && p->eEvalmode==FTS3_EVAL_NEXT)
2878 );
2879
2880 if( 0==(rc = fts3EvalExpr(p, pExpr->pRight, &aRight, &nRight, isReqPos))
2881 && 0==(rc = fts3EvalExpr(p, pExpr->pLeft, &aLeft, &nLeft, isReqPos))
2882 ){
2883 switch( pExpr->eType ){
2884 case FTSQUERY_NEAR: {
2885 Fts3Expr *pLeft;
2886 Fts3Expr *pRight;
2887 int mergetype = MERGE_NEAR;
2888 if( pExpr->pParent && pExpr->pParent->eType==FTSQUERY_NEAR ){
2889 mergetype = MERGE_POS_NEAR;
2890 }
2891 pLeft = pExpr->pLeft;
2892 while( pLeft->eType==FTSQUERY_NEAR ){
2893 pLeft=pLeft->pRight;
2894 }
2895 pRight = pExpr->pRight;
2896 assert( pRight->eType==FTSQUERY_PHRASE );
2897 assert( pLeft->eType==FTSQUERY_PHRASE );
2898
2899 rc = fts3NearMerge(mergetype, pExpr->nNear,
2900 pLeft->pPhrase->nToken, aLeft, nLeft,
2901 pRight->pPhrase->nToken, aRight, nRight,
2902 paOut, pnOut
2903 );
2904 sqlite3_free(aLeft);
2905 break;
2906 }
2907
2908 case FTSQUERY_OR: {
2909 /* Allocate a buffer for the output. The maximum size is the
2910 ** sum of the sizes of the two input buffers. The +1 term is
2911 ** so that a buffer of zero bytes is never allocated - this can
2912 ** cause fts3DoclistMerge() to incorrectly return SQLITE_NOMEM.
2913 */
2914 char *aBuffer = sqlite3_malloc(nRight+nLeft+1);
2915 rc = fts3DoclistMerge(MERGE_OR, 0, 0, aBuffer, pnOut,
2916 aLeft, nLeft, aRight, nRight, 0
2917 );
2918 *paOut = aBuffer;
2919 sqlite3_free(aLeft);
2920 break;
2921 }
2922
2923 default: {
2924 assert( FTSQUERY_NOT==MERGE_NOT && FTSQUERY_AND==MERGE_AND );
2925 fts3DoclistMerge(pExpr->eType, 0, 0, aLeft, pnOut,
2926 aLeft, nLeft, aRight, nRight, 0
2927 );
2928 *paOut = aLeft;
2929 break;
2930 }
2931 }
2932 }
2933 sqlite3_free(aRight);
2934 }
2935 }
2936
2937 assert( rc==SQLITE_OK || *paOut==0 );
2938 return rc;
2939 }
2940
2941 /*
2942 ** This function is called from within xNext() for each row visited by
2943 ** an FTS3 query. If evaluating the FTS3 query expression within xFilter()
2944 ** was able to determine the exact set of matching rows, this function sets
2945 ** *pbRes to true and returns SQLITE_IO immediately.
2946 **
2947 ** Otherwise, if evaluating the query expression within xFilter() returned a
2948 ** superset of the matching documents instead of an exact set (this happens
2949 ** when the query includes very common tokens and it is deemed too expensive to
2950 ** load their doclists from disk), this function tests if the current row
2951 ** really does match the FTS3 query.
2952 **
2953 ** If an error occurs, an SQLite error code is returned. Otherwise, SQLITE_OK
2954 ** is returned and *pbRes is set to true if the current row matches the
2955 ** FTS3 query (and should be included in the results returned to SQLite), or
2956 ** false otherwise.
2957 */
2958 static int fts3EvalDeferred(
2959 Fts3Cursor *pCsr, /* FTS3 cursor pointing at row to test */
2960 int *pbRes /* OUT: Set to true if row is a match */
2961 ){
2962 int rc = SQLITE_OK;
2963 if( pCsr->pDeferred==0 ){
2964 *pbRes = 1;
2965 }else{
2966 rc = fts3CursorSeek(0, pCsr);
2967 if( rc==SQLITE_OK ){
2968 sqlite3Fts3FreeDeferredDoclists(pCsr);
2969 rc = sqlite3Fts3CacheDeferredDoclists(pCsr);
2970 }
2971 if( rc==SQLITE_OK ){
2972 char *a = 0;
2973 int n = 0;
2974 rc = fts3EvalExpr(pCsr, pCsr->pExpr, &a, &n, 0);
2975 assert( n>=0 );
2976 *pbRes = (n>0);
2977 sqlite3_free(a);
2978 }
2979 }
2980 return rc;
2981 }
2982
2983 /*
2984 ** Advance the cursor to the next row in the %_content table that 3018 ** Advance the cursor to the next row in the %_content table that
2985 ** matches the search criteria. For a MATCH search, this will be 3019 ** matches the search criteria. For a MATCH search, this will be
2986 ** the next row that matches. For a full-table scan, this will be 3020 ** the next row that matches. For a full-table scan, this will be
2987 ** simply the next row in the %_content table. For a docid lookup, 3021 ** simply the next row in the %_content table. For a docid lookup,
2988 ** this routine simply sets the EOF flag. 3022 ** this routine simply sets the EOF flag.
2989 ** 3023 **
2990 ** Return SQLITE_OK if nothing goes wrong. SQLITE_OK is returned 3024 ** Return SQLITE_OK if nothing goes wrong. SQLITE_OK is returned
2991 ** even if we reach end-of-file. The fts3EofMethod() will be called 3025 ** even if we reach end-of-file. The fts3EofMethod() will be called
2992 ** subsequently to determine whether or not an EOF was hit. 3026 ** subsequently to determine whether or not an EOF was hit.
2993 */ 3027 */
2994 static int fts3NextMethod(sqlite3_vtab_cursor *pCursor){ 3028 static int fts3NextMethod(sqlite3_vtab_cursor *pCursor){
2995 int res; 3029 int rc;
2996 int rc = SQLITE_OK; /* Return code */
2997 Fts3Cursor *pCsr = (Fts3Cursor *)pCursor; 3030 Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;
2998 3031 if( pCsr->eSearch==FTS3_DOCID_SEARCH || pCsr->eSearch==FTS3_FULLSCAN_SEARCH ){
2999 pCsr->eEvalmode = FTS3_EVAL_NEXT; 3032 if( SQLITE_ROW!=sqlite3_step(pCsr->pStmt) ){
3000 do { 3033 pCsr->isEof = 1;
3001 if( pCsr->aDoclist==0 ){ 3034 rc = sqlite3_reset(pCsr->pStmt);
3002 if( SQLITE_ROW!=sqlite3_step(pCsr->pStmt) ){ 3035 }else{
3003 pCsr->isEof = 1;
3004 rc = sqlite3_reset(pCsr->pStmt);
3005 break;
3006 }
3007 pCsr->iPrevId = sqlite3_column_int64(pCsr->pStmt, 0); 3036 pCsr->iPrevId = sqlite3_column_int64(pCsr->pStmt, 0);
3008 }else{ 3037 rc = SQLITE_OK;
3009 if( pCsr->pNextId>=&pCsr->aDoclist[pCsr->nDoclist] ){ 3038 }
3010 pCsr->isEof = 1; 3039 }else{
3011 break; 3040 rc = fts3EvalNext((Fts3Cursor *)pCursor);
3012 } 3041 }
3013 sqlite3_reset(pCsr->pStmt); 3042 assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
3014 fts3GetDeltaVarint(&pCsr->pNextId, &pCsr->iPrevId);
3015 pCsr->isRequireSeek = 1;
3016 pCsr->isMatchinfoNeeded = 1;
3017 }
3018 }while( SQLITE_OK==(rc = fts3EvalDeferred(pCsr, &res)) && res==0 );
3019
3020 return rc; 3043 return rc;
3021 } 3044 }
3022 3045
3023 /* 3046 /*
3047 ** The following are copied from sqliteInt.h.
3048 **
3049 ** Constants for the largest and smallest possible 64-bit signed integers.
3050 ** These macros are designed to work correctly on both 32-bit and 64-bit
3051 ** compilers.
3052 */
3053 #ifndef SQLITE_AMALGAMATION
3054 # define LARGEST_INT64 (0xffffffff|(((sqlite3_int64)0x7fffffff)<<32))
3055 # define SMALLEST_INT64 (((sqlite3_int64)-1) - LARGEST_INT64)
3056 #endif
3057
3058 /*
3059 ** If the numeric type of argument pVal is "integer", then return it
3060 ** converted to a 64-bit signed integer. Otherwise, return a copy of
3061 ** the second parameter, iDefault.
3062 */
3063 static sqlite3_int64 fts3DocidRange(sqlite3_value *pVal, i64 iDefault){
3064 if( pVal ){
3065 int eType = sqlite3_value_numeric_type(pVal);
3066 if( eType==SQLITE_INTEGER ){
3067 return sqlite3_value_int64(pVal);
3068 }
3069 }
3070 return iDefault;
3071 }
3072
3073 /*
3024 ** This is the xFilter interface for the virtual table. See 3074 ** This is the xFilter interface for the virtual table. See
3025 ** the virtual table xFilter method documentation for additional 3075 ** the virtual table xFilter method documentation for additional
3026 ** information. 3076 ** information.
3027 ** 3077 **
3028 ** If idxNum==FTS3_FULLSCAN_SEARCH then do a full table scan against 3078 ** If idxNum==FTS3_FULLSCAN_SEARCH then do a full table scan against
3029 ** the %_content table. 3079 ** the %_content table.
3030 ** 3080 **
3031 ** If idxNum==FTS3_DOCID_SEARCH then do a docid lookup for a single entry 3081 ** If idxNum==FTS3_DOCID_SEARCH then do a docid lookup for a single entry
3032 ** in the %_content table. 3082 ** in the %_content table.
3033 ** 3083 **
3034 ** If idxNum>=FTS3_FULLTEXT_SEARCH then use the full text index. The 3084 ** If idxNum>=FTS3_FULLTEXT_SEARCH then use the full text index. The
3035 ** column on the left-hand side of the MATCH operator is column 3085 ** column on the left-hand side of the MATCH operator is column
3036 ** number idxNum-FTS3_FULLTEXT_SEARCH, 0 indexed. argv[0] is the right-hand 3086 ** number idxNum-FTS3_FULLTEXT_SEARCH, 0 indexed. argv[0] is the right-hand
3037 ** side of the MATCH operator. 3087 ** side of the MATCH operator.
3038 */ 3088 */
3039 static int fts3FilterMethod( 3089 static int fts3FilterMethod(
3040 sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */ 3090 sqlite3_vtab_cursor *pCursor, /* The cursor used for this query */
3041 int idxNum, /* Strategy index */ 3091 int idxNum, /* Strategy index */
3042 const char *idxStr, /* Unused */ 3092 const char *idxStr, /* Unused */
3043 int nVal, /* Number of elements in apVal */ 3093 int nVal, /* Number of elements in apVal */
3044 sqlite3_value **apVal /* Arguments for the indexing scheme */ 3094 sqlite3_value **apVal /* Arguments for the indexing scheme */
3045 ){ 3095 ){
3046 const char *azSql[] = { 3096 int rc;
3047 "SELECT %s FROM %Q.'%q_content' AS x WHERE docid = ?", /* non-full-scan */
3048 "SELECT %s FROM %Q.'%q_content' AS x ", /* full-scan */
3049 };
3050 int rc; /* Return code */
3051 char *zSql; /* SQL statement used to access %_content */ 3097 char *zSql; /* SQL statement used to access %_content */
3098 int eSearch;
3052 Fts3Table *p = (Fts3Table *)pCursor->pVtab; 3099 Fts3Table *p = (Fts3Table *)pCursor->pVtab;
3053 Fts3Cursor *pCsr = (Fts3Cursor *)pCursor; 3100 Fts3Cursor *pCsr = (Fts3Cursor *)pCursor;
3054 3101
3102 sqlite3_value *pCons = 0; /* The MATCH or rowid constraint, if any */
3103 sqlite3_value *pLangid = 0; /* The "langid = ?" constraint, if any */
3104 sqlite3_value *pDocidGe = 0; /* The "docid >= ?" constraint, if any */
3105 sqlite3_value *pDocidLe = 0; /* The "docid <= ?" constraint, if any */
3106 int iIdx;
3107
3055 UNUSED_PARAMETER(idxStr); 3108 UNUSED_PARAMETER(idxStr);
3056 UNUSED_PARAMETER(nVal); 3109 UNUSED_PARAMETER(nVal);
3057 3110
3058 assert( idxNum>=0 && idxNum<=(FTS3_FULLTEXT_SEARCH+p->nColumn) ); 3111 eSearch = (idxNum & 0x0000FFFF);
3059 assert( nVal==0 || nVal==1 ); 3112 assert( eSearch>=0 && eSearch<=(FTS3_FULLTEXT_SEARCH+p->nColumn) );
3060 assert( (nVal==0)==(idxNum==FTS3_FULLSCAN_SEARCH) );
3061 assert( p->pSegments==0 ); 3113 assert( p->pSegments==0 );
3062 3114
3115 /* Collect arguments into local variables */
3116 iIdx = 0;
3117 if( eSearch!=FTS3_FULLSCAN_SEARCH ) pCons = apVal[iIdx++];
3118 if( idxNum & FTS3_HAVE_LANGID ) pLangid = apVal[iIdx++];
3119 if( idxNum & FTS3_HAVE_DOCID_GE ) pDocidGe = apVal[iIdx++];
3120 if( idxNum & FTS3_HAVE_DOCID_LE ) pDocidLe = apVal[iIdx++];
3121 assert( iIdx==nVal );
3122
3063 /* In case the cursor has been used before, clear it now. */ 3123 /* In case the cursor has been used before, clear it now. */
3064 sqlite3_finalize(pCsr->pStmt); 3124 sqlite3_finalize(pCsr->pStmt);
3065 sqlite3_free(pCsr->aDoclist); 3125 sqlite3_free(pCsr->aDoclist);
3126 sqlite3_free(pCsr->aMatchinfo);
3066 sqlite3Fts3ExprFree(pCsr->pExpr); 3127 sqlite3Fts3ExprFree(pCsr->pExpr);
3067 memset(&pCursor[1], 0, sizeof(Fts3Cursor)-sizeof(sqlite3_vtab_cursor)); 3128 memset(&pCursor[1], 0, sizeof(Fts3Cursor)-sizeof(sqlite3_vtab_cursor));
3068 3129
3069 if( idxNum!=FTS3_DOCID_SEARCH && idxNum!=FTS3_FULLSCAN_SEARCH ){ 3130 /* Set the lower and upper bounds on docids to return */
3070 int iCol = idxNum-FTS3_FULLTEXT_SEARCH; 3131 pCsr->iMinDocid = fts3DocidRange(pDocidGe, SMALLEST_INT64);
3071 const char *zQuery = (const char *)sqlite3_value_text(apVal[0]); 3132 pCsr->iMaxDocid = fts3DocidRange(pDocidLe, LARGEST_INT64);
3072 3133
3073 if( zQuery==0 && sqlite3_value_type(apVal[0])!=SQLITE_NULL ){ 3134 if( idxStr ){
3135 pCsr->bDesc = (idxStr[0]=='D');
3136 }else{
3137 pCsr->bDesc = p->bDescIdx;
3138 }
3139 pCsr->eSearch = (i16)eSearch;
3140
3141 if( eSearch!=FTS3_DOCID_SEARCH && eSearch!=FTS3_FULLSCAN_SEARCH ){
3142 int iCol = eSearch-FTS3_FULLTEXT_SEARCH;
3143 const char *zQuery = (const char *)sqlite3_value_text(pCons);
3144
3145 if( zQuery==0 && sqlite3_value_type(pCons)!=SQLITE_NULL ){
3074 return SQLITE_NOMEM; 3146 return SQLITE_NOMEM;
3075 } 3147 }
3076 3148
3077 rc = sqlite3Fts3ExprParse(p->pTokenizer, p->azColumn, p->nColumn, 3149 pCsr->iLangid = 0;
3078 iCol, zQuery, -1, &pCsr->pExpr 3150 if( pLangid ) pCsr->iLangid = sqlite3_value_int(pLangid);
3151
3152 assert( p->base.zErrMsg==0 );
3153 rc = sqlite3Fts3ExprParse(p->pTokenizer, pCsr->iLangid,
3154 p->azColumn, p->bFts4, p->nColumn, iCol, zQuery, -1, &pCsr->pExpr,
3155 &p->base.zErrMsg
3079 ); 3156 );
3080 if( rc!=SQLITE_OK ){ 3157 if( rc!=SQLITE_OK ){
3081 if( rc==SQLITE_ERROR ){
3082 p->base.zErrMsg = sqlite3_mprintf("malformed MATCH expression: [%s]",
3083 zQuery);
3084 }
3085 return rc; 3158 return rc;
3086 } 3159 }
3087 3160
3088 rc = sqlite3Fts3ReadLock(p); 3161 rc = fts3EvalStart(pCsr);
3089 if( rc!=SQLITE_OK ) return rc;
3090
3091 rc = fts3EvalExpr(pCsr, pCsr->pExpr, &pCsr->aDoclist, &pCsr->nDoclist, 0);
3092 sqlite3Fts3SegmentsClose(p); 3162 sqlite3Fts3SegmentsClose(p);
3093 if( rc!=SQLITE_OK ) return rc; 3163 if( rc!=SQLITE_OK ) return rc;
3094 pCsr->pNextId = pCsr->aDoclist; 3164 pCsr->pNextId = pCsr->aDoclist;
3095 pCsr->iPrevId = 0; 3165 pCsr->iPrevId = 0;
3096 } 3166 }
3097 3167
3098 /* Compile a SELECT statement for this cursor. For a full-table-scan, the 3168 /* Compile a SELECT statement for this cursor. For a full-table-scan, the
3099 ** statement loops through all rows of the %_content table. For a 3169 ** statement loops through all rows of the %_content table. For a
3100 ** full-text query or docid lookup, the statement retrieves a single 3170 ** full-text query or docid lookup, the statement retrieves a single
3101 ** row by docid. 3171 ** row by docid.
3102 */ 3172 */
3103 zSql = (char *)azSql[idxNum==FTS3_FULLSCAN_SEARCH]; 3173 if( eSearch==FTS3_FULLSCAN_SEARCH ){
3104 zSql = sqlite3_mprintf(zSql, p->zReadExprlist, p->zDb, p->zName); 3174 zSql = sqlite3_mprintf(
3105 if( !zSql ){ 3175 "SELECT %s ORDER BY rowid %s",
3106 rc = SQLITE_NOMEM; 3176 p->zReadExprlist, (pCsr->bDesc ? "DESC" : "ASC")
3107 }else{ 3177 );
3108 rc = sqlite3_prepare_v2(p->db, zSql, -1, &pCsr->pStmt, 0); 3178 if( zSql ){
3109 sqlite3_free(zSql); 3179 rc = sqlite3_prepare_v2(p->db, zSql, -1, &pCsr->pStmt, 0);
3110 } 3180 sqlite3_free(zSql);
3111 if( rc==SQLITE_OK && idxNum==FTS3_DOCID_SEARCH ){ 3181 }else{
3112 rc = sqlite3_bind_value(pCsr->pStmt, 1, apVal[0]); 3182 rc = SQLITE_NOMEM;
3113 } 3183 }
3114 pCsr->eSearch = (i16)idxNum; 3184 }else if( eSearch==FTS3_DOCID_SEARCH ){
3115 3185 rc = fts3CursorSeekStmt(pCsr, &pCsr->pStmt);
3186 if( rc==SQLITE_OK ){
3187 rc = sqlite3_bind_value(pCsr->pStmt, 1, pCons);
3188 }
3189 }
3116 if( rc!=SQLITE_OK ) return rc; 3190 if( rc!=SQLITE_OK ) return rc;
3191
3117 return fts3NextMethod(pCursor); 3192 return fts3NextMethod(pCursor);
3118 } 3193 }
3119 3194
3120 /* 3195 /*
3121 ** This is the xEof method of the virtual table. SQLite calls this 3196 ** This is the xEof method of the virtual table. SQLite calls this
3122 ** routine to find out if it has reached the end of a result set. 3197 ** routine to find out if it has reached the end of a result set.
3123 */ 3198 */
3124 static int fts3EofMethod(sqlite3_vtab_cursor *pCursor){ 3199 static int fts3EofMethod(sqlite3_vtab_cursor *pCursor){
3125 return ((Fts3Cursor *)pCursor)->isEof; 3200 return ((Fts3Cursor *)pCursor)->isEof;
3126 } 3201 }
3127 3202
3128 /* 3203 /*
3129 ** This is the xRowid method. The SQLite core calls this routine to 3204 ** This is the xRowid method. The SQLite core calls this routine to
3130 ** retrieve the rowid for the current row of the result set. fts3 3205 ** retrieve the rowid for the current row of the result set. fts3
3131 ** exposes %_content.docid as the rowid for the virtual table. The 3206 ** exposes %_content.docid as the rowid for the virtual table. The
3132 ** rowid should be written to *pRowid. 3207 ** rowid should be written to *pRowid.
3133 */ 3208 */
3134 static int fts3RowidMethod(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){ 3209 static int fts3RowidMethod(sqlite3_vtab_cursor *pCursor, sqlite_int64 *pRowid){
3135 Fts3Cursor *pCsr = (Fts3Cursor *) pCursor; 3210 Fts3Cursor *pCsr = (Fts3Cursor *) pCursor;
3136 if( pCsr->aDoclist ){ 3211 *pRowid = pCsr->iPrevId;
3137 *pRowid = pCsr->iPrevId;
3138 }else{
3139 /* This branch runs if the query is implemented using a full-table scan
3140 ** (not using the full-text index). In this case grab the rowid from the
3141 ** SELECT statement.
3142 */
3143 assert( pCsr->isRequireSeek==0 );
3144 *pRowid = sqlite3_column_int64(pCsr->pStmt, 0);
3145 }
3146 return SQLITE_OK; 3212 return SQLITE_OK;
3147 } 3213 }
3148 3214
3149 /* 3215 /*
3150 ** This is the xColumn method, called by SQLite to request a value from 3216 ** This is the xColumn method, called by SQLite to request a value from
3151 ** the row that the supplied cursor currently points to. 3217 ** the row that the supplied cursor currently points to.
3218 **
3219 ** If:
3220 **
3221 ** (iCol < p->nColumn) -> The value of the iCol'th user column.
3222 ** (iCol == p->nColumn) -> Magic column with the same name as the table.
3223 ** (iCol == p->nColumn+1) -> Docid column
3224 ** (iCol == p->nColumn+2) -> Langid column
3152 */ 3225 */
3153 static int fts3ColumnMethod( 3226 static int fts3ColumnMethod(
3154 sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */ 3227 sqlite3_vtab_cursor *pCursor, /* Cursor to retrieve value from */
3155 sqlite3_context *pContext, /* Context for sqlite3_result_xxx() calls */ 3228 sqlite3_context *pCtx, /* Context for sqlite3_result_xxx() calls */
3156 int iCol /* Index of column to read value from */ 3229 int iCol /* Index of column to read value from */
3157 ){ 3230 ){
3158 int rc; /* Return Code */ 3231 int rc = SQLITE_OK; /* Return Code */
3159 Fts3Cursor *pCsr = (Fts3Cursor *) pCursor; 3232 Fts3Cursor *pCsr = (Fts3Cursor *) pCursor;
3160 Fts3Table *p = (Fts3Table *)pCursor->pVtab; 3233 Fts3Table *p = (Fts3Table *)pCursor->pVtab;
3161 3234
3162 /* The column value supplied by SQLite must be in range. */ 3235 /* The column value supplied by SQLite must be in range. */
3163 assert( iCol>=0 && iCol<=p->nColumn+1 ); 3236 assert( iCol>=0 && iCol<=p->nColumn+2 );
3164 3237
3165 if( iCol==p->nColumn+1 ){ 3238 if( iCol==p->nColumn+1 ){
3166 /* This call is a request for the "docid" column. Since "docid" is an 3239 /* This call is a request for the "docid" column. Since "docid" is an
3167 ** alias for "rowid", use the xRowid() method to obtain the value. 3240 ** alias for "rowid", use the xRowid() method to obtain the value.
3168 */ 3241 */
3169 sqlite3_int64 iRowid; 3242 sqlite3_result_int64(pCtx, pCsr->iPrevId);
3170 rc = fts3RowidMethod(pCursor, &iRowid);
3171 sqlite3_result_int64(pContext, iRowid);
3172 }else if( iCol==p->nColumn ){ 3243 }else if( iCol==p->nColumn ){
3173 /* The extra column whose name is the same as the table. 3244 /* The extra column whose name is the same as the table.
3174 ** Return a blob which is a pointer to the cursor. 3245 ** Return a blob which is a pointer to the cursor. */
3175 */ 3246 sqlite3_result_blob(pCtx, &pCsr, sizeof(pCsr), SQLITE_TRANSIENT);
3176 sqlite3_result_blob(pContext, &pCsr, sizeof(pCsr), SQLITE_TRANSIENT); 3247 }else if( iCol==p->nColumn+2 && pCsr->pExpr ){
3177 rc = SQLITE_OK; 3248 sqlite3_result_int64(pCtx, pCsr->iLangid);
3178 }else{ 3249 }else{
3250 /* The requested column is either a user column (one that contains
3251 ** indexed data), or the language-id column. */
3179 rc = fts3CursorSeek(0, pCsr); 3252 rc = fts3CursorSeek(0, pCsr);
3253
3180 if( rc==SQLITE_OK ){ 3254 if( rc==SQLITE_OK ){
3181 sqlite3_result_value(pContext, sqlite3_column_value(pCsr->pStmt, iCol+1)); 3255 if( iCol==p->nColumn+2 ){
3182 } 3256 int iLangid = 0;
3183 } 3257 if( p->zLanguageid ){
3258 iLangid = sqlite3_column_int(pCsr->pStmt, p->nColumn+1);
3259 }
3260 sqlite3_result_int(pCtx, iLangid);
3261 }else if( sqlite3_data_count(pCsr->pStmt)>(iCol+1) ){
3262 sqlite3_result_value(pCtx, sqlite3_column_value(pCsr->pStmt, iCol+1));
3263 }
3264 }
3265 }
3266
3267 assert( ((Fts3Table *)pCsr->base.pVtab)->pSegments==0 );
3184 return rc; 3268 return rc;
3185 } 3269 }
3186 3270
3187 /* 3271 /*
3188 ** This function is the implementation of the xUpdate callback used by 3272 ** This function is the implementation of the xUpdate callback used by
3189 ** FTS3 virtual tables. It is invoked by SQLite each time a row is to be 3273 ** FTS3 virtual tables. It is invoked by SQLite each time a row is to be
3190 ** inserted, updated or deleted. 3274 ** inserted, updated or deleted.
3191 */ 3275 */
3192 static int fts3UpdateMethod( 3276 static int fts3UpdateMethod(
3193 sqlite3_vtab *pVtab, /* Virtual table handle */ 3277 sqlite3_vtab *pVtab, /* Virtual table handle */
3194 int nArg, /* Size of argument array */ 3278 int nArg, /* Size of argument array */
3195 sqlite3_value **apVal, /* Array of arguments */ 3279 sqlite3_value **apVal, /* Array of arguments */
3196 sqlite_int64 *pRowid /* OUT: The affected (or effected) rowid */ 3280 sqlite_int64 *pRowid /* OUT: The affected (or effected) rowid */
3197 ){ 3281 ){
3198 return sqlite3Fts3UpdateMethod(pVtab, nArg, apVal, pRowid); 3282 return sqlite3Fts3UpdateMethod(pVtab, nArg, apVal, pRowid);
3199 } 3283 }
3200 3284
3201 /* 3285 /*
3202 ** Implementation of xSync() method. Flush the contents of the pending-terms 3286 ** Implementation of xSync() method. Flush the contents of the pending-terms
3203 ** hash-table to the database. 3287 ** hash-table to the database.
3204 */ 3288 */
3205 static int fts3SyncMethod(sqlite3_vtab *pVtab){ 3289 static int fts3SyncMethod(sqlite3_vtab *pVtab){
3206 int rc = sqlite3Fts3PendingTermsFlush((Fts3Table *)pVtab); 3290
3207 sqlite3Fts3SegmentsClose((Fts3Table *)pVtab); 3291 /* Following an incremental-merge operation, assuming that the input
3292 ** segments are not completely consumed (the usual case), they are updated
3293 ** in place to remove the entries that have already been merged. This
3294 ** involves updating the leaf block that contains the smallest unmerged
3295 ** entry and each block (if any) between the leaf and the root node. So
3296 ** if the height of the input segment b-trees is N, and input segments
3297 ** are merged eight at a time, updating the input segments at the end
3298 ** of an incremental-merge requires writing (8*(1+N)) blocks. N is usually
3299 ** small - often between 0 and 2. So the overhead of the incremental
3300 ** merge is somewhere between 8 and 24 blocks. To avoid this overhead
3301 ** dwarfing the actual productive work accomplished, the incremental merge
3302 ** is only attempted if it will write at least 64 leaf blocks. Hence
3303 ** nMinMerge.
3304 **
3305 ** Of course, updating the input segments also involves deleting a bunch
3306 ** of blocks from the segments table. But this is not considered overhead
3307 ** as it would also be required by a crisis-merge that used the same input
3308 ** segments.
3309 */
3310 const u32 nMinMerge = 64; /* Minimum amount of incr-merge work to do */
3311
3312 Fts3Table *p = (Fts3Table*)pVtab;
3313 int rc = sqlite3Fts3PendingTermsFlush(p);
3314
3315 if( rc==SQLITE_OK
3316 && p->nLeafAdd>(nMinMerge/16)
3317 && p->nAutoincrmerge && p->nAutoincrmerge!=0xff
3318 ){
3319 int mxLevel = 0; /* Maximum relative level value in db */
3320 int A; /* Incr-merge parameter A */
3321
3322 rc = sqlite3Fts3MaxLevel(p, &mxLevel);
3323 assert( rc==SQLITE_OK || mxLevel==0 );
3324 A = p->nLeafAdd * mxLevel;
3325 A += (A/2);
3326 if( A>(int)nMinMerge ) rc = sqlite3Fts3Incrmerge(p, A, p->nAutoincrmerge);
3327 }
3328 sqlite3Fts3SegmentsClose(p);
3208 return rc; 3329 return rc;
3209 } 3330 }
3210 3331
3211 /* 3332 /*
3212 ** Implementation of xBegin() method. This is a no-op. 3333 ** If it is currently unknown whether or not the FTS table has an %_stat
3334 ** table (if p->bHasStat==2), attempt to determine this (set p->bHasStat
3335 ** to 0 or 1). Return SQLITE_OK if successful, or an SQLite error code
3336 ** if an error occurs.
3337 */
3338 static int fts3SetHasStat(Fts3Table *p){
3339 int rc = SQLITE_OK;
3340 if( p->bHasStat==2 ){
3341 const char *zFmt ="SELECT 1 FROM %Q.sqlite_master WHERE tbl_name='%q_stat'";
3342 char *zSql = sqlite3_mprintf(zFmt, p->zDb, p->zName);
3343 if( zSql ){
3344 sqlite3_stmt *pStmt = 0;
3345 rc = sqlite3_prepare_v2(p->db, zSql, -1, &pStmt, 0);
3346 if( rc==SQLITE_OK ){
3347 int bHasStat = (sqlite3_step(pStmt)==SQLITE_ROW);
3348 rc = sqlite3_finalize(pStmt);
3349 if( rc==SQLITE_OK ) p->bHasStat = bHasStat;
3350 }
3351 sqlite3_free(zSql);
3352 }else{
3353 rc = SQLITE_NOMEM;
3354 }
3355 }
3356 return rc;
3357 }
3358
3359 /*
3360 ** Implementation of xBegin() method.
3213 */ 3361 */
3214 static int fts3BeginMethod(sqlite3_vtab *pVtab){ 3362 static int fts3BeginMethod(sqlite3_vtab *pVtab){
3363 Fts3Table *p = (Fts3Table*)pVtab;
3215 UNUSED_PARAMETER(pVtab); 3364 UNUSED_PARAMETER(pVtab);
3216 assert( ((Fts3Table *)pVtab)->nPendingData==0 ); 3365 assert( p->pSegments==0 );
3217 return SQLITE_OK; 3366 assert( p->nPendingData==0 );
3367 assert( p->inTransaction!=1 );
3368 TESTONLY( p->inTransaction = 1 );
3369 TESTONLY( p->mxSavepoint = -1; );
3370 p->nLeafAdd = 0;
3371 return fts3SetHasStat(p);
3218 } 3372 }
3219 3373
3220 /* 3374 /*
3221 ** Implementation of xCommit() method. This is a no-op. The contents of 3375 ** Implementation of xCommit() method. This is a no-op. The contents of
3222 ** the pending-terms hash-table have already been flushed into the database 3376 ** the pending-terms hash-table have already been flushed into the database
3223 ** by fts3SyncMethod(). 3377 ** by fts3SyncMethod().
3224 */ 3378 */
3225 static int fts3CommitMethod(sqlite3_vtab *pVtab){ 3379 static int fts3CommitMethod(sqlite3_vtab *pVtab){
3380 TESTONLY( Fts3Table *p = (Fts3Table*)pVtab );
3226 UNUSED_PARAMETER(pVtab); 3381 UNUSED_PARAMETER(pVtab);
3227 assert( ((Fts3Table *)pVtab)->nPendingData==0 ); 3382 assert( p->nPendingData==0 );
3383 assert( p->inTransaction!=0 );
3384 assert( p->pSegments==0 );
3385 TESTONLY( p->inTransaction = 0 );
3386 TESTONLY( p->mxSavepoint = -1; );
3228 return SQLITE_OK; 3387 return SQLITE_OK;
3229 } 3388 }
3230 3389
3231 /* 3390 /*
3232 ** Implementation of xRollback(). Discard the contents of the pending-terms 3391 ** Implementation of xRollback(). Discard the contents of the pending-terms
3233 ** hash-table. Any changes made to the database are reverted by SQLite. 3392 ** hash-table. Any changes made to the database are reverted by SQLite.
3234 */ 3393 */
3235 static int fts3RollbackMethod(sqlite3_vtab *pVtab){ 3394 static int fts3RollbackMethod(sqlite3_vtab *pVtab){
3236 sqlite3Fts3PendingTermsClear((Fts3Table *)pVtab); 3395 Fts3Table *p = (Fts3Table*)pVtab;
3396 sqlite3Fts3PendingTermsClear(p);
3397 assert( p->inTransaction!=0 );
3398 TESTONLY( p->inTransaction = 0 );
3399 TESTONLY( p->mxSavepoint = -1; );
3237 return SQLITE_OK; 3400 return SQLITE_OK;
3238 } 3401 }
3239 3402
3240 /* 3403 /*
3241 ** Load the doclist associated with expression pExpr to pExpr->aDoclist. 3404 ** When called, *ppPoslist must point to the byte immediately following the
3242 ** The loaded doclist contains positions as well as the document ids. 3405 ** end of a position-list. i.e. ( (*ppPoslist)[-1]==POS_END ). This function
3243 ** This is used by the matchinfo(), snippet() and offsets() auxillary 3406 ** moves *ppPoslist so that it instead points to the first byte of the
3244 ** functions. 3407 ** same position list.
3245 */ 3408 */
3246 int sqlite3Fts3ExprLoadDoclist(Fts3Cursor *pCsr, Fts3Expr *pExpr){ 3409 static void fts3ReversePoslist(char *pStart, char **ppPoslist){
3247 int rc; 3410 char *p = &(*ppPoslist)[-2];
3248 assert( pExpr->eType==FTSQUERY_PHRASE && pExpr->pPhrase ); 3411 char c = 0;
3249 assert( pCsr->eEvalmode==FTS3_EVAL_NEXT ); 3412
3250 rc = fts3EvalExpr(pCsr, pExpr, &pExpr->aDoclist, &pExpr->nDoclist, 1); 3413 while( p>pStart && (c=*p--)==0 );
3251 return rc; 3414 while( p>pStart && (*p & 0x80) | c ){
3252 } 3415 c = *p--;
3253 3416 }
3254 int sqlite3Fts3ExprLoadFtDoclist( 3417 if( p>pStart ){ p = &p[2]; }
3255 Fts3Cursor *pCsr, 3418 while( *p++&0x80 );
3256 Fts3Expr *pExpr, 3419 *ppPoslist = p;
3257 char **paDoclist,
3258 int *pnDoclist
3259 ){
3260 int rc;
3261 assert( pCsr->eEvalmode==FTS3_EVAL_NEXT );
3262 assert( pExpr->eType==FTSQUERY_PHRASE && pExpr->pPhrase );
3263 pCsr->eEvalmode = FTS3_EVAL_MATCHINFO;
3264 rc = fts3EvalExpr(pCsr, pExpr, paDoclist, pnDoclist, 1);
3265 pCsr->eEvalmode = FTS3_EVAL_NEXT;
3266 return rc;
3267 }
3268
3269 /*
3270 ** After ExprLoadDoclist() (see above) has been called, this function is
3271 ** used to iterate/search through the position lists that make up the doclist
3272 ** stored in pExpr->aDoclist.
3273 */
3274 char *sqlite3Fts3FindPositions(
3275 Fts3Expr *pExpr, /* Access this expressions doclist */
3276 sqlite3_int64 iDocid, /* Docid associated with requested pos-list */
3277 int iCol /* Column of requested pos-list */
3278 ){
3279 assert( pExpr->isLoaded );
3280 if( pExpr->aDoclist ){
3281 char *pEnd = &pExpr->aDoclist[pExpr->nDoclist];
3282 char *pCsr;
3283
3284 if( pExpr->pCurrent==0 ){
3285 pExpr->pCurrent = pExpr->aDoclist;
3286 pExpr->iCurrent = 0;
3287 pExpr->pCurrent += sqlite3Fts3GetVarint(pExpr->pCurrent,&pExpr->iCurrent);
3288 }
3289 pCsr = pExpr->pCurrent;
3290 assert( pCsr );
3291
3292 while( pCsr<pEnd ){
3293 if( pExpr->iCurrent<iDocid ){
3294 fts3PoslistCopy(0, &pCsr);
3295 if( pCsr<pEnd ){
3296 fts3GetDeltaVarint(&pCsr, &pExpr->iCurrent);
3297 }
3298 pExpr->pCurrent = pCsr;
3299 }else{
3300 if( pExpr->iCurrent==iDocid ){
3301 int iThis = 0;
3302 if( iCol<0 ){
3303 /* If iCol is negative, return a pointer to the start of the
3304 ** position-list (instead of a pointer to the start of a list
3305 ** of offsets associated with a specific column).
3306 */
3307 return pCsr;
3308 }
3309 while( iThis<iCol ){
3310 fts3ColumnlistCopy(0, &pCsr);
3311 if( *pCsr==0x00 ) return 0;
3312 pCsr++;
3313 pCsr += sqlite3Fts3GetVarint32(pCsr, &iThis);
3314 }
3315 if( iCol==iThis && (*pCsr&0xFE) ) return pCsr;
3316 }
3317 return 0;
3318 }
3319 }
3320 }
3321
3322 return 0;
3323 } 3420 }
3324 3421
3325 /* 3422 /*
3326 ** Helper function used by the implementation of the overloaded snippet(), 3423 ** Helper function used by the implementation of the overloaded snippet(),
3327 ** offsets() and optimize() SQL functions. 3424 ** offsets() and optimize() SQL functions.
3328 ** 3425 **
3329 ** If the value passed as the third argument is a blob of size 3426 ** If the value passed as the third argument is a blob of size
3330 ** sizeof(Fts3Cursor*), then the blob contents are copied to the 3427 ** sizeof(Fts3Cursor*), then the blob contents are copied to the
3331 ** output variable *ppCsr and SQLITE_OK is returned. Otherwise, an error 3428 ** output variable *ppCsr and SQLITE_OK is returned. Otherwise, an error
3332 ** message is written to context pContext and SQLITE_ERROR returned. The 3429 ** message is written to context pContext and SQLITE_ERROR returned. The
(...skipping 180 matching lines...) Expand 10 before | Expand all | Expand 10 after
3513 ** Implementation of FTS3 xRename method. Rename an fts3 table. 3610 ** Implementation of FTS3 xRename method. Rename an fts3 table.
3514 */ 3611 */
3515 static int fts3RenameMethod( 3612 static int fts3RenameMethod(
3516 sqlite3_vtab *pVtab, /* Virtual table handle */ 3613 sqlite3_vtab *pVtab, /* Virtual table handle */
3517 const char *zName /* New name of table */ 3614 const char *zName /* New name of table */
3518 ){ 3615 ){
3519 Fts3Table *p = (Fts3Table *)pVtab; 3616 Fts3Table *p = (Fts3Table *)pVtab;
3520 sqlite3 *db = p->db; /* Database connection */ 3617 sqlite3 *db = p->db; /* Database connection */
3521 int rc; /* Return Code */ 3618 int rc; /* Return Code */
3522 3619
3523 rc = sqlite3Fts3PendingTermsFlush(p); 3620 /* At this point it must be known if the %_stat table exists or not.
3524 if( rc!=SQLITE_OK ){ 3621 ** So bHasStat may not be 2. */
3525 return rc; 3622 rc = fts3SetHasStat(p);
3623
3624 /* As it happens, the pending terms table is always empty here. This is
3625 ** because an "ALTER TABLE RENAME TABLE" statement inside a transaction
3626 ** always opens a savepoint transaction. And the xSavepoint() method
3627 ** flushes the pending terms table. But leave the (no-op) call to
3628 ** PendingTermsFlush() in in case that changes.
3629 */
3630 assert( p->nPendingData==0 );
3631 if( rc==SQLITE_OK ){
3632 rc = sqlite3Fts3PendingTermsFlush(p);
3526 } 3633 }
3527 3634
3528 fts3DbExec(&rc, db, 3635 if( p->zContentTbl==0 ){
3529 "ALTER TABLE %Q.'%q_content' RENAME TO '%q_content';", 3636 fts3DbExec(&rc, db,
3530 p->zDb, p->zName, zName 3637 "ALTER TABLE %Q.'%q_content' RENAME TO '%q_content';",
3531 ); 3638 p->zDb, p->zName, zName
3639 );
3640 }
3641
3532 if( p->bHasDocsize ){ 3642 if( p->bHasDocsize ){
3533 fts3DbExec(&rc, db, 3643 fts3DbExec(&rc, db,
3534 "ALTER TABLE %Q.'%q_docsize' RENAME TO '%q_docsize';", 3644 "ALTER TABLE %Q.'%q_docsize' RENAME TO '%q_docsize';",
3535 p->zDb, p->zName, zName 3645 p->zDb, p->zName, zName
3536 ); 3646 );
3537 } 3647 }
3538 if( p->bHasStat ){ 3648 if( p->bHasStat ){
3539 fts3DbExec(&rc, db, 3649 fts3DbExec(&rc, db,
3540 "ALTER TABLE %Q.'%q_stat' RENAME TO '%q_stat';", 3650 "ALTER TABLE %Q.'%q_stat' RENAME TO '%q_stat';",
3541 p->zDb, p->zName, zName 3651 p->zDb, p->zName, zName
3542 ); 3652 );
3543 } 3653 }
3544 fts3DbExec(&rc, db, 3654 fts3DbExec(&rc, db,
3545 "ALTER TABLE %Q.'%q_segments' RENAME TO '%q_segments';", 3655 "ALTER TABLE %Q.'%q_segments' RENAME TO '%q_segments';",
3546 p->zDb, p->zName, zName 3656 p->zDb, p->zName, zName
3547 ); 3657 );
3548 fts3DbExec(&rc, db, 3658 fts3DbExec(&rc, db,
3549 "ALTER TABLE %Q.'%q_segdir' RENAME TO '%q_segdir';", 3659 "ALTER TABLE %Q.'%q_segdir' RENAME TO '%q_segdir';",
3550 p->zDb, p->zName, zName 3660 p->zDb, p->zName, zName
3551 ); 3661 );
3552 return rc; 3662 return rc;
3553 } 3663 }
3554 3664
3665 /*
3666 ** The xSavepoint() method.
3667 **
3668 ** Flush the contents of the pending-terms table to disk.
3669 */
3670 static int fts3SavepointMethod(sqlite3_vtab *pVtab, int iSavepoint){
3671 int rc = SQLITE_OK;
3672 UNUSED_PARAMETER(iSavepoint);
3673 assert( ((Fts3Table *)pVtab)->inTransaction );
3674 assert( ((Fts3Table *)pVtab)->mxSavepoint < iSavepoint );
3675 TESTONLY( ((Fts3Table *)pVtab)->mxSavepoint = iSavepoint );
3676 if( ((Fts3Table *)pVtab)->bIgnoreSavepoint==0 ){
3677 rc = fts3SyncMethod(pVtab);
3678 }
3679 return rc;
3680 }
3681
3682 /*
3683 ** The xRelease() method.
3684 **
3685 ** This is a no-op.
3686 */
3687 static int fts3ReleaseMethod(sqlite3_vtab *pVtab, int iSavepoint){
3688 TESTONLY( Fts3Table *p = (Fts3Table*)pVtab );
3689 UNUSED_PARAMETER(iSavepoint);
3690 UNUSED_PARAMETER(pVtab);
3691 assert( p->inTransaction );
3692 assert( p->mxSavepoint >= iSavepoint );
3693 TESTONLY( p->mxSavepoint = iSavepoint-1 );
3694 return SQLITE_OK;
3695 }
3696
3697 /*
3698 ** The xRollbackTo() method.
3699 **
3700 ** Discard the contents of the pending terms table.
3701 */
3702 static int fts3RollbackToMethod(sqlite3_vtab *pVtab, int iSavepoint){
3703 Fts3Table *p = (Fts3Table*)pVtab;
3704 UNUSED_PARAMETER(iSavepoint);
3705 assert( p->inTransaction );
3706 assert( p->mxSavepoint >= iSavepoint );
3707 TESTONLY( p->mxSavepoint = iSavepoint );
3708 sqlite3Fts3PendingTermsClear(p);
3709 return SQLITE_OK;
3710 }
3711
3555 static const sqlite3_module fts3Module = { 3712 static const sqlite3_module fts3Module = {
3556 /* iVersion */ 0, 3713 /* iVersion */ 2,
3557 /* xCreate */ fts3CreateMethod, 3714 /* xCreate */ fts3CreateMethod,
3558 /* xConnect */ fts3ConnectMethod, 3715 /* xConnect */ fts3ConnectMethod,
3559 /* xBestIndex */ fts3BestIndexMethod, 3716 /* xBestIndex */ fts3BestIndexMethod,
3560 /* xDisconnect */ fts3DisconnectMethod, 3717 /* xDisconnect */ fts3DisconnectMethod,
3561 /* xDestroy */ fts3DestroyMethod, 3718 /* xDestroy */ fts3DestroyMethod,
3562 /* xOpen */ fts3OpenMethod, 3719 /* xOpen */ fts3OpenMethod,
3563 /* xClose */ fts3CloseMethod, 3720 /* xClose */ fts3CloseMethod,
3564 /* xFilter */ fts3FilterMethod, 3721 /* xFilter */ fts3FilterMethod,
3565 /* xNext */ fts3NextMethod, 3722 /* xNext */ fts3NextMethod,
3566 /* xEof */ fts3EofMethod, 3723 /* xEof */ fts3EofMethod,
3567 /* xColumn */ fts3ColumnMethod, 3724 /* xColumn */ fts3ColumnMethod,
3568 /* xRowid */ fts3RowidMethod, 3725 /* xRowid */ fts3RowidMethod,
3569 /* xUpdate */ fts3UpdateMethod, 3726 /* xUpdate */ fts3UpdateMethod,
3570 /* xBegin */ fts3BeginMethod, 3727 /* xBegin */ fts3BeginMethod,
3571 /* xSync */ fts3SyncMethod, 3728 /* xSync */ fts3SyncMethod,
3572 /* xCommit */ fts3CommitMethod, 3729 /* xCommit */ fts3CommitMethod,
3573 /* xRollback */ fts3RollbackMethod, 3730 /* xRollback */ fts3RollbackMethod,
3574 /* xFindFunction */ fts3FindFunctionMethod, 3731 /* xFindFunction */ fts3FindFunctionMethod,
3575 /* xRename */ fts3RenameMethod, 3732 /* xRename */ fts3RenameMethod,
3733 /* xSavepoint */ fts3SavepointMethod,
3734 /* xRelease */ fts3ReleaseMethod,
3735 /* xRollbackTo */ fts3RollbackToMethod,
3576 }; 3736 };
3577 3737
3578 /* 3738 /*
3579 ** This function is registered as the module destructor (called when an 3739 ** This function is registered as the module destructor (called when an
3580 ** FTS3 enabled database connection is closed). It frees the memory 3740 ** FTS3 enabled database connection is closed). It frees the memory
3581 ** allocated for the tokenizer hash table. 3741 ** allocated for the tokenizer hash table.
3582 */ 3742 */
3583 static void hashDestroy(void *p){ 3743 static void hashDestroy(void *p){
3584 Fts3Hash *pHash = (Fts3Hash *)p; 3744 Fts3Hash *pHash = (Fts3Hash *)p;
3585 sqlite3Fts3HashClear(pHash); 3745 sqlite3Fts3HashClear(pHash);
3586 sqlite3_free(pHash); 3746 sqlite3_free(pHash);
3587 } 3747 }
3588 3748
3589 /* 3749 /*
3590 ** The fts3 built-in tokenizers - "simple", "porter" and "icu"- are 3750 ** The fts3 built-in tokenizers - "simple", "porter" and "icu"- are
3591 ** implemented in files fts3_tokenizer1.c, fts3_porter.c and fts3_icu.c 3751 ** implemented in files fts3_tokenizer1.c, fts3_porter.c and fts3_icu.c
3592 ** respectively. The following three forward declarations are for functions 3752 ** respectively. The following three forward declarations are for functions
3593 ** declared in these files used to retrieve the respective implementations. 3753 ** declared in these files used to retrieve the respective implementations.
3594 ** 3754 **
3595 ** Calling sqlite3Fts3SimpleTokenizerModule() sets the value pointed 3755 ** Calling sqlite3Fts3SimpleTokenizerModule() sets the value pointed
3596 ** to by the argument to point to the "simple" tokenizer implementation. 3756 ** to by the argument to point to the "simple" tokenizer implementation.
3597 ** And so on. 3757 ** And so on.
3598 */ 3758 */
3599 void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule); 3759 void sqlite3Fts3SimpleTokenizerModule(sqlite3_tokenizer_module const**ppModule);
3600 void sqlite3Fts3PorterTokenizerModule(sqlite3_tokenizer_module const**ppModule); 3760 void sqlite3Fts3PorterTokenizerModule(sqlite3_tokenizer_module const**ppModule);
3761 #ifndef SQLITE_DISABLE_FTS3_UNICODE
3762 void sqlite3Fts3UnicodeTokenizer(sqlite3_tokenizer_module const**ppModule);
3763 #endif
3601 #ifdef SQLITE_ENABLE_ICU 3764 #ifdef SQLITE_ENABLE_ICU
3602 void sqlite3Fts3IcuTokenizerModule(sqlite3_tokenizer_module const**ppModule); 3765 void sqlite3Fts3IcuTokenizerModule(sqlite3_tokenizer_module const**ppModule);
3603 #endif 3766 #endif
3604 3767
3605 /* 3768 /*
3606 ** Initialise the fts3 extension. If this extension is built as part 3769 ** Initialize the fts3 extension. If this extension is built as part
3607 ** of the sqlite library, then this function is called directly by 3770 ** of the sqlite library, then this function is called directly by
3608 ** SQLite. If fts3 is built as a dynamically loadable extension, this 3771 ** SQLite. If fts3 is built as a dynamically loadable extension, this
3609 ** function is called by the sqlite3_extension_init() entry point. 3772 ** function is called by the sqlite3_extension_init() entry point.
3610 */ 3773 */
3611 int sqlite3Fts3Init(sqlite3 *db){ 3774 int sqlite3Fts3Init(sqlite3 *db){
3612 int rc = SQLITE_OK; 3775 int rc = SQLITE_OK;
3613 Fts3Hash *pHash = 0; 3776 Fts3Hash *pHash = 0;
3614 const sqlite3_tokenizer_module *pSimple = 0; 3777 const sqlite3_tokenizer_module *pSimple = 0;
3615 const sqlite3_tokenizer_module *pPorter = 0; 3778 const sqlite3_tokenizer_module *pPorter = 0;
3779 #ifndef SQLITE_DISABLE_FTS3_UNICODE
3780 const sqlite3_tokenizer_module *pUnicode = 0;
3781 #endif
3616 3782
3617 #ifdef SQLITE_ENABLE_ICU 3783 #ifdef SQLITE_ENABLE_ICU
3618 const sqlite3_tokenizer_module *pIcu = 0; 3784 const sqlite3_tokenizer_module *pIcu = 0;
3619 sqlite3Fts3IcuTokenizerModule(&pIcu); 3785 sqlite3Fts3IcuTokenizerModule(&pIcu);
3620 #endif 3786 #endif
3621 3787
3788 #ifndef SQLITE_DISABLE_FTS3_UNICODE
3789 sqlite3Fts3UnicodeTokenizer(&pUnicode);
3790 #endif
3791
3792 #ifdef SQLITE_TEST
3793 rc = sqlite3Fts3InitTerm(db);
3794 if( rc!=SQLITE_OK ) return rc;
3795 #endif
3796
3622 rc = sqlite3Fts3InitAux(db); 3797 rc = sqlite3Fts3InitAux(db);
3623 if( rc!=SQLITE_OK ) return rc; 3798 if( rc!=SQLITE_OK ) return rc;
3624 3799
3625 sqlite3Fts3SimpleTokenizerModule(&pSimple); 3800 sqlite3Fts3SimpleTokenizerModule(&pSimple);
3626 sqlite3Fts3PorterTokenizerModule(&pPorter); 3801 sqlite3Fts3PorterTokenizerModule(&pPorter);
3627 3802
3628 /* Allocate and initialise the hash-table used to store tokenizers. */ 3803 /* Allocate and initialize the hash-table used to store tokenizers. */
3629 pHash = sqlite3_malloc(sizeof(Fts3Hash)); 3804 pHash = sqlite3_malloc(sizeof(Fts3Hash));
3630 if( !pHash ){ 3805 if( !pHash ){
3631 rc = SQLITE_NOMEM; 3806 rc = SQLITE_NOMEM;
3632 }else{ 3807 }else{
3633 sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1); 3808 sqlite3Fts3HashInit(pHash, FTS3_HASH_STRING, 1);
3634 } 3809 }
3635 3810
3636 /* Load the built-in tokenizers into the hash table */ 3811 /* Load the built-in tokenizers into the hash table */
3637 if( rc==SQLITE_OK ){ 3812 if( rc==SQLITE_OK ){
3638 if( sqlite3Fts3HashInsert(pHash, "simple", 7, (void *)pSimple) 3813 if( sqlite3Fts3HashInsert(pHash, "simple", 7, (void *)pSimple)
3639 || sqlite3Fts3HashInsert(pHash, "porter", 7, (void *)pPorter) 3814 || sqlite3Fts3HashInsert(pHash, "porter", 7, (void *)pPorter)
3815
3816 #ifndef SQLITE_DISABLE_FTS3_UNICODE
3817 || sqlite3Fts3HashInsert(pHash, "unicode61", 10, (void *)pUnicode)
3818 #endif
3640 #ifdef SQLITE_ENABLE_ICU 3819 #ifdef SQLITE_ENABLE_ICU
3641 || (pIcu && sqlite3Fts3HashInsert(pHash, "icu", 4, (void *)pIcu)) 3820 || (pIcu && sqlite3Fts3HashInsert(pHash, "icu", 4, (void *)pIcu))
3642 #endif 3821 #endif
3643 ){ 3822 ){
3644 rc = SQLITE_NOMEM; 3823 rc = SQLITE_NOMEM;
3645 } 3824 }
3646 } 3825 }
3647 3826
3648 #ifdef SQLITE_TEST 3827 #ifdef SQLITE_TEST
3649 if( rc==SQLITE_OK ){ 3828 if( rc==SQLITE_OK ){
(...skipping 14 matching lines...) Expand all
3664 && SQLITE_OK==(rc = sqlite3_overload_function(db, "snippet", -1)) 3843 && SQLITE_OK==(rc = sqlite3_overload_function(db, "snippet", -1))
3665 && SQLITE_OK==(rc = sqlite3_overload_function(db, "offsets", 1)) 3844 && SQLITE_OK==(rc = sqlite3_overload_function(db, "offsets", 1))
3666 && SQLITE_OK==(rc = sqlite3_overload_function(db, "matchinfo", 1)) 3845 && SQLITE_OK==(rc = sqlite3_overload_function(db, "matchinfo", 1))
3667 && SQLITE_OK==(rc = sqlite3_overload_function(db, "matchinfo", 2)) 3846 && SQLITE_OK==(rc = sqlite3_overload_function(db, "matchinfo", 2))
3668 && SQLITE_OK==(rc = sqlite3_overload_function(db, "optimize", 1)) 3847 && SQLITE_OK==(rc = sqlite3_overload_function(db, "optimize", 1))
3669 ){ 3848 ){
3670 rc = sqlite3_create_module_v2( 3849 rc = sqlite3_create_module_v2(
3671 db, "fts3", &fts3Module, (void *)pHash, hashDestroy 3850 db, "fts3", &fts3Module, (void *)pHash, hashDestroy
3672 ); 3851 );
3673 #if CHROMIUM_FTS3_CHANGES && !SQLITE_TEST 3852 #if CHROMIUM_FTS3_CHANGES && !SQLITE_TEST
3674 /* Disable fts4 pending review. */ 3853 /* Disable fts4 and tokenizer vtab pending review. */
3675 #else 3854 #else
3676 if( rc==SQLITE_OK ){ 3855 if( rc==SQLITE_OK ){
3677 rc = sqlite3_create_module_v2( 3856 rc = sqlite3_create_module_v2(
3678 db, "fts4", &fts3Module, (void *)pHash, 0 3857 db, "fts4", &fts3Module, (void *)pHash, 0
3679 ); 3858 );
3680 } 3859 }
3860 if( rc==SQLITE_OK ){
3861 rc = sqlite3Fts3InitTok(db, (void *)pHash);
3862 }
3681 #endif 3863 #endif
3682 return rc; 3864 return rc;
3683 } 3865 }
3684 3866
3867
3685 /* An error has occurred. Delete the hash table and return the error code. */ 3868 /* An error has occurred. Delete the hash table and return the error code. */
3686 assert( rc!=SQLITE_OK ); 3869 assert( rc!=SQLITE_OK );
3687 if( pHash ){ 3870 if( pHash ){
3688 sqlite3Fts3HashClear(pHash); 3871 sqlite3Fts3HashClear(pHash);
3689 sqlite3_free(pHash); 3872 sqlite3_free(pHash);
3690 } 3873 }
3691 return rc; 3874 return rc;
3692 } 3875 }
3693 3876
3877 /*
3878 ** Allocate an Fts3MultiSegReader for each token in the expression headed
3879 ** by pExpr.
3880 **
3881 ** An Fts3SegReader object is a cursor that can seek or scan a range of
3882 ** entries within a single segment b-tree. An Fts3MultiSegReader uses multiple
3883 ** Fts3SegReader objects internally to provide an interface to seek or scan
3884 ** within the union of all segments of a b-tree. Hence the name.
3885 **
3886 ** If the allocated Fts3MultiSegReader just seeks to a single entry in a
3887 ** segment b-tree (if the term is not a prefix or it is a prefix for which
3888 ** there exists prefix b-tree of the right length) then it may be traversed
3889 ** and merged incrementally. Otherwise, it has to be merged into an in-memory
3890 ** doclist and then traversed.
3891 */
3892 static void fts3EvalAllocateReaders(
3893 Fts3Cursor *pCsr, /* FTS cursor handle */
3894 Fts3Expr *pExpr, /* Allocate readers for this expression */
3895 int *pnToken, /* OUT: Total number of tokens in phrase. */
3896 int *pnOr, /* OUT: Total number of OR nodes in expr. */
3897 int *pRc /* IN/OUT: Error code */
3898 ){
3899 if( pExpr && SQLITE_OK==*pRc ){
3900 if( pExpr->eType==FTSQUERY_PHRASE ){
3901 int i;
3902 int nToken = pExpr->pPhrase->nToken;
3903 *pnToken += nToken;
3904 for(i=0; i<nToken; i++){
3905 Fts3PhraseToken *pToken = &pExpr->pPhrase->aToken[i];
3906 int rc = fts3TermSegReaderCursor(pCsr,
3907 pToken->z, pToken->n, pToken->isPrefix, &pToken->pSegcsr
3908 );
3909 if( rc!=SQLITE_OK ){
3910 *pRc = rc;
3911 return;
3912 }
3913 }
3914 assert( pExpr->pPhrase->iDoclistToken==0 );
3915 pExpr->pPhrase->iDoclistToken = -1;
3916 }else{
3917 *pnOr += (pExpr->eType==FTSQUERY_OR);
3918 fts3EvalAllocateReaders(pCsr, pExpr->pLeft, pnToken, pnOr, pRc);
3919 fts3EvalAllocateReaders(pCsr, pExpr->pRight, pnToken, pnOr, pRc);
3920 }
3921 }
3922 }
3923
3924 /*
3925 ** Arguments pList/nList contain the doclist for token iToken of phrase p.
3926 ** It is merged into the main doclist stored in p->doclist.aAll/nAll.
3927 **
3928 ** This function assumes that pList points to a buffer allocated using
3929 ** sqlite3_malloc(). This function takes responsibility for eventually
3930 ** freeing the buffer.
3931 */
3932 static void fts3EvalPhraseMergeToken(
3933 Fts3Table *pTab, /* FTS Table pointer */
3934 Fts3Phrase *p, /* Phrase to merge pList/nList into */
3935 int iToken, /* Token pList/nList corresponds to */
3936 char *pList, /* Pointer to doclist */
3937 int nList /* Number of bytes in pList */
3938 ){
3939 assert( iToken!=p->iDoclistToken );
3940
3941 if( pList==0 ){
3942 sqlite3_free(p->doclist.aAll);
3943 p->doclist.aAll = 0;
3944 p->doclist.nAll = 0;
3945 }
3946
3947 else if( p->iDoclistToken<0 ){
3948 p->doclist.aAll = pList;
3949 p->doclist.nAll = nList;
3950 }
3951
3952 else if( p->doclist.aAll==0 ){
3953 sqlite3_free(pList);
3954 }
3955
3956 else {
3957 char *pLeft;
3958 char *pRight;
3959 int nLeft;
3960 int nRight;
3961 int nDiff;
3962
3963 if( p->iDoclistToken<iToken ){
3964 pLeft = p->doclist.aAll;
3965 nLeft = p->doclist.nAll;
3966 pRight = pList;
3967 nRight = nList;
3968 nDiff = iToken - p->iDoclistToken;
3969 }else{
3970 pRight = p->doclist.aAll;
3971 nRight = p->doclist.nAll;
3972 pLeft = pList;
3973 nLeft = nList;
3974 nDiff = p->iDoclistToken - iToken;
3975 }
3976
3977 fts3DoclistPhraseMerge(pTab->bDescIdx, nDiff, pLeft, nLeft, pRight,&nRight);
3978 sqlite3_free(pLeft);
3979 p->doclist.aAll = pRight;
3980 p->doclist.nAll = nRight;
3981 }
3982
3983 if( iToken>p->iDoclistToken ) p->iDoclistToken = iToken;
3984 }
3985
3986 /*
3987 ** Load the doclist for phrase p into p->doclist.aAll/nAll. The loaded doclist
3988 ** does not take deferred tokens into account.
3989 **
3990 ** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
3991 */
3992 static int fts3EvalPhraseLoad(
3993 Fts3Cursor *pCsr, /* FTS Cursor handle */
3994 Fts3Phrase *p /* Phrase object */
3995 ){
3996 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
3997 int iToken;
3998 int rc = SQLITE_OK;
3999
4000 for(iToken=0; rc==SQLITE_OK && iToken<p->nToken; iToken++){
4001 Fts3PhraseToken *pToken = &p->aToken[iToken];
4002 assert( pToken->pDeferred==0 || pToken->pSegcsr==0 );
4003
4004 if( pToken->pSegcsr ){
4005 int nThis = 0;
4006 char *pThis = 0;
4007 rc = fts3TermSelect(pTab, pToken, p->iColumn, &nThis, &pThis);
4008 if( rc==SQLITE_OK ){
4009 fts3EvalPhraseMergeToken(pTab, p, iToken, pThis, nThis);
4010 }
4011 }
4012 assert( pToken->pSegcsr==0 );
4013 }
4014
4015 return rc;
4016 }
4017
4018 /*
4019 ** This function is called on each phrase after the position lists for
4020 ** any deferred tokens have been loaded into memory. It updates the phrases
4021 ** current position list to include only those positions that are really
4022 ** instances of the phrase (after considering deferred tokens). If this
4023 ** means that the phrase does not appear in the current row, doclist.pList
4024 ** and doclist.nList are both zeroed.
4025 **
4026 ** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
4027 */
4028 static int fts3EvalDeferredPhrase(Fts3Cursor *pCsr, Fts3Phrase *pPhrase){
4029 int iToken; /* Used to iterate through phrase tokens */
4030 char *aPoslist = 0; /* Position list for deferred tokens */
4031 int nPoslist = 0; /* Number of bytes in aPoslist */
4032 int iPrev = -1; /* Token number of previous deferred token */
4033
4034 assert( pPhrase->doclist.bFreeList==0 );
4035
4036 for(iToken=0; iToken<pPhrase->nToken; iToken++){
4037 Fts3PhraseToken *pToken = &pPhrase->aToken[iToken];
4038 Fts3DeferredToken *pDeferred = pToken->pDeferred;
4039
4040 if( pDeferred ){
4041 char *pList;
4042 int nList;
4043 int rc = sqlite3Fts3DeferredTokenList(pDeferred, &pList, &nList);
4044 if( rc!=SQLITE_OK ) return rc;
4045
4046 if( pList==0 ){
4047 sqlite3_free(aPoslist);
4048 pPhrase->doclist.pList = 0;
4049 pPhrase->doclist.nList = 0;
4050 return SQLITE_OK;
4051
4052 }else if( aPoslist==0 ){
4053 aPoslist = pList;
4054 nPoslist = nList;
4055
4056 }else{
4057 char *aOut = pList;
4058 char *p1 = aPoslist;
4059 char *p2 = aOut;
4060
4061 assert( iPrev>=0 );
4062 fts3PoslistPhraseMerge(&aOut, iToken-iPrev, 0, 1, &p1, &p2);
4063 sqlite3_free(aPoslist);
4064 aPoslist = pList;
4065 nPoslist = (int)(aOut - aPoslist);
4066 if( nPoslist==0 ){
4067 sqlite3_free(aPoslist);
4068 pPhrase->doclist.pList = 0;
4069 pPhrase->doclist.nList = 0;
4070 return SQLITE_OK;
4071 }
4072 }
4073 iPrev = iToken;
4074 }
4075 }
4076
4077 if( iPrev>=0 ){
4078 int nMaxUndeferred = pPhrase->iDoclistToken;
4079 if( nMaxUndeferred<0 ){
4080 pPhrase->doclist.pList = aPoslist;
4081 pPhrase->doclist.nList = nPoslist;
4082 pPhrase->doclist.iDocid = pCsr->iPrevId;
4083 pPhrase->doclist.bFreeList = 1;
4084 }else{
4085 int nDistance;
4086 char *p1;
4087 char *p2;
4088 char *aOut;
4089
4090 if( nMaxUndeferred>iPrev ){
4091 p1 = aPoslist;
4092 p2 = pPhrase->doclist.pList;
4093 nDistance = nMaxUndeferred - iPrev;
4094 }else{
4095 p1 = pPhrase->doclist.pList;
4096 p2 = aPoslist;
4097 nDistance = iPrev - nMaxUndeferred;
4098 }
4099
4100 aOut = (char *)sqlite3_malloc(nPoslist+8);
4101 if( !aOut ){
4102 sqlite3_free(aPoslist);
4103 return SQLITE_NOMEM;
4104 }
4105
4106 pPhrase->doclist.pList = aOut;
4107 if( fts3PoslistPhraseMerge(&aOut, nDistance, 0, 1, &p1, &p2) ){
4108 pPhrase->doclist.bFreeList = 1;
4109 pPhrase->doclist.nList = (int)(aOut - pPhrase->doclist.pList);
4110 }else{
4111 sqlite3_free(aOut);
4112 pPhrase->doclist.pList = 0;
4113 pPhrase->doclist.nList = 0;
4114 }
4115 sqlite3_free(aPoslist);
4116 }
4117 }
4118
4119 return SQLITE_OK;
4120 }
4121
4122 /*
4123 ** Maximum number of tokens a phrase may have to be considered for the
4124 ** incremental doclists strategy.
4125 */
4126 #define MAX_INCR_PHRASE_TOKENS 4
4127
4128 /*
4129 ** This function is called for each Fts3Phrase in a full-text query
4130 ** expression to initialize the mechanism for returning rows. Once this
4131 ** function has been called successfully on an Fts3Phrase, it may be
4132 ** used with fts3EvalPhraseNext() to iterate through the matching docids.
4133 **
4134 ** If parameter bOptOk is true, then the phrase may (or may not) use the
4135 ** incremental loading strategy. Otherwise, the entire doclist is loaded into
4136 ** memory within this call.
4137 **
4138 ** SQLITE_OK is returned if no error occurs, otherwise an SQLite error code.
4139 */
4140 static int fts3EvalPhraseStart(Fts3Cursor *pCsr, int bOptOk, Fts3Phrase *p){
4141 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
4142 int rc = SQLITE_OK; /* Error code */
4143 int i;
4144
4145 /* Determine if doclists may be loaded from disk incrementally. This is
4146 ** possible if the bOptOk argument is true, the FTS doclists will be
4147 ** scanned in forward order, and the phrase consists of
4148 ** MAX_INCR_PHRASE_TOKENS or fewer tokens, none of which are are "^first"
4149 ** tokens or prefix tokens that cannot use a prefix-index. */
4150 int bHaveIncr = 0;
4151 int bIncrOk = (bOptOk
4152 && pCsr->bDesc==pTab->bDescIdx
4153 && p->nToken<=MAX_INCR_PHRASE_TOKENS && p->nToken>0
4154 && p->nToken<=MAX_INCR_PHRASE_TOKENS && p->nToken>0
4155 #ifdef SQLITE_TEST
4156 && pTab->bNoIncrDoclist==0
4157 #endif
4158 );
4159 for(i=0; bIncrOk==1 && i<p->nToken; i++){
4160 Fts3PhraseToken *pToken = &p->aToken[i];
4161 if( pToken->bFirst || (pToken->pSegcsr!=0 && !pToken->pSegcsr->bLookup) ){
4162 bIncrOk = 0;
4163 }
4164 if( pToken->pSegcsr ) bHaveIncr = 1;
4165 }
4166
4167 if( bIncrOk && bHaveIncr ){
4168 /* Use the incremental approach. */
4169 int iCol = (p->iColumn >= pTab->nColumn ? -1 : p->iColumn);
4170 for(i=0; rc==SQLITE_OK && i<p->nToken; i++){
4171 Fts3PhraseToken *pToken = &p->aToken[i];
4172 Fts3MultiSegReader *pSegcsr = pToken->pSegcsr;
4173 if( pSegcsr ){
4174 rc = sqlite3Fts3MsrIncrStart(pTab, pSegcsr, iCol, pToken->z, pToken->n);
4175 }
4176 }
4177 p->bIncr = 1;
4178 }else{
4179 /* Load the full doclist for the phrase into memory. */
4180 rc = fts3EvalPhraseLoad(pCsr, p);
4181 p->bIncr = 0;
4182 }
4183
4184 assert( rc!=SQLITE_OK || p->nToken<1 || p->aToken[0].pSegcsr==0 || p->bIncr );
4185 return rc;
4186 }
4187
4188 /*
4189 ** This function is used to iterate backwards (from the end to start)
4190 ** through doclists. It is used by this module to iterate through phrase
4191 ** doclists in reverse and by the fts3_write.c module to iterate through
4192 ** pending-terms lists when writing to databases with "order=desc".
4193 **
4194 ** The doclist may be sorted in ascending (parameter bDescIdx==0) or
4195 ** descending (parameter bDescIdx==1) order of docid. Regardless, this
4196 ** function iterates from the end of the doclist to the beginning.
4197 */
4198 void sqlite3Fts3DoclistPrev(
4199 int bDescIdx, /* True if the doclist is desc */
4200 char *aDoclist, /* Pointer to entire doclist */
4201 int nDoclist, /* Length of aDoclist in bytes */
4202 char **ppIter, /* IN/OUT: Iterator pointer */
4203 sqlite3_int64 *piDocid, /* IN/OUT: Docid pointer */
4204 int *pnList, /* OUT: List length pointer */
4205 u8 *pbEof /* OUT: End-of-file flag */
4206 ){
4207 char *p = *ppIter;
4208
4209 assert( nDoclist>0 );
4210 assert( *pbEof==0 );
4211 assert( p || *piDocid==0 );
4212 assert( !p || (p>aDoclist && p<&aDoclist[nDoclist]) );
4213
4214 if( p==0 ){
4215 sqlite3_int64 iDocid = 0;
4216 char *pNext = 0;
4217 char *pDocid = aDoclist;
4218 char *pEnd = &aDoclist[nDoclist];
4219 int iMul = 1;
4220
4221 while( pDocid<pEnd ){
4222 sqlite3_int64 iDelta;
4223 pDocid += sqlite3Fts3GetVarint(pDocid, &iDelta);
4224 iDocid += (iMul * iDelta);
4225 pNext = pDocid;
4226 fts3PoslistCopy(0, &pDocid);
4227 while( pDocid<pEnd && *pDocid==0 ) pDocid++;
4228 iMul = (bDescIdx ? -1 : 1);
4229 }
4230
4231 *pnList = (int)(pEnd - pNext);
4232 *ppIter = pNext;
4233 *piDocid = iDocid;
4234 }else{
4235 int iMul = (bDescIdx ? -1 : 1);
4236 sqlite3_int64 iDelta;
4237 fts3GetReverseVarint(&p, aDoclist, &iDelta);
4238 *piDocid -= (iMul * iDelta);
4239
4240 if( p==aDoclist ){
4241 *pbEof = 1;
4242 }else{
4243 char *pSave = p;
4244 fts3ReversePoslist(aDoclist, &p);
4245 *pnList = (int)(pSave - p);
4246 }
4247 *ppIter = p;
4248 }
4249 }
4250
4251 /*
4252 ** Iterate forwards through a doclist.
4253 */
4254 void sqlite3Fts3DoclistNext(
4255 int bDescIdx, /* True if the doclist is desc */
4256 char *aDoclist, /* Pointer to entire doclist */
4257 int nDoclist, /* Length of aDoclist in bytes */
4258 char **ppIter, /* IN/OUT: Iterator pointer */
4259 sqlite3_int64 *piDocid, /* IN/OUT: Docid pointer */
4260 u8 *pbEof /* OUT: End-of-file flag */
4261 ){
4262 char *p = *ppIter;
4263
4264 assert( nDoclist>0 );
4265 assert( *pbEof==0 );
4266 assert( p || *piDocid==0 );
4267 assert( !p || (p>=aDoclist && p<=&aDoclist[nDoclist]) );
4268
4269 if( p==0 ){
4270 p = aDoclist;
4271 p += sqlite3Fts3GetVarint(p, piDocid);
4272 }else{
4273 fts3PoslistCopy(0, &p);
4274 if( p>=&aDoclist[nDoclist] ){
4275 *pbEof = 1;
4276 }else{
4277 sqlite3_int64 iVar;
4278 p += sqlite3Fts3GetVarint(p, &iVar);
4279 *piDocid += ((bDescIdx ? -1 : 1) * iVar);
4280 }
4281 }
4282
4283 *ppIter = p;
4284 }
4285
4286 /*
4287 ** Advance the iterator pDL to the next entry in pDL->aAll/nAll. Set *pbEof
4288 ** to true if EOF is reached.
4289 */
4290 static void fts3EvalDlPhraseNext(
4291 Fts3Table *pTab,
4292 Fts3Doclist *pDL,
4293 u8 *pbEof
4294 ){
4295 char *pIter; /* Used to iterate through aAll */
4296 char *pEnd = &pDL->aAll[pDL->nAll]; /* 1 byte past end of aAll */
4297
4298 if( pDL->pNextDocid ){
4299 pIter = pDL->pNextDocid;
4300 }else{
4301 pIter = pDL->aAll;
4302 }
4303
4304 if( pIter>=pEnd ){
4305 /* We have already reached the end of this doclist. EOF. */
4306 *pbEof = 1;
4307 }else{
4308 sqlite3_int64 iDelta;
4309 pIter += sqlite3Fts3GetVarint(pIter, &iDelta);
4310 if( pTab->bDescIdx==0 || pDL->pNextDocid==0 ){
4311 pDL->iDocid += iDelta;
4312 }else{
4313 pDL->iDocid -= iDelta;
4314 }
4315 pDL->pList = pIter;
4316 fts3PoslistCopy(0, &pIter);
4317 pDL->nList = (int)(pIter - pDL->pList);
4318
4319 /* pIter now points just past the 0x00 that terminates the position-
4320 ** list for document pDL->iDocid. However, if this position-list was
4321 ** edited in place by fts3EvalNearTrim(), then pIter may not actually
4322 ** point to the start of the next docid value. The following line deals
4323 ** with this case by advancing pIter past the zero-padding added by
4324 ** fts3EvalNearTrim(). */
4325 while( pIter<pEnd && *pIter==0 ) pIter++;
4326
4327 pDL->pNextDocid = pIter;
4328 assert( pIter>=&pDL->aAll[pDL->nAll] || *pIter );
4329 *pbEof = 0;
4330 }
4331 }
4332
4333 /*
4334 ** Helper type used by fts3EvalIncrPhraseNext() and incrPhraseTokenNext().
4335 */
4336 typedef struct TokenDoclist TokenDoclist;
4337 struct TokenDoclist {
4338 int bIgnore;
4339 sqlite3_int64 iDocid;
4340 char *pList;
4341 int nList;
4342 };
4343
4344 /*
4345 ** Token pToken is an incrementally loaded token that is part of a
4346 ** multi-token phrase. Advance it to the next matching document in the
4347 ** database and populate output variable *p with the details of the new
4348 ** entry. Or, if the iterator has reached EOF, set *pbEof to true.
4349 **
4350 ** If an error occurs, return an SQLite error code. Otherwise, return
4351 ** SQLITE_OK.
4352 */
4353 static int incrPhraseTokenNext(
4354 Fts3Table *pTab, /* Virtual table handle */
4355 Fts3Phrase *pPhrase, /* Phrase to advance token of */
4356 int iToken, /* Specific token to advance */
4357 TokenDoclist *p, /* OUT: Docid and doclist for new entry */
4358 u8 *pbEof /* OUT: True if iterator is at EOF */
4359 ){
4360 int rc = SQLITE_OK;
4361
4362 if( pPhrase->iDoclistToken==iToken ){
4363 assert( p->bIgnore==0 );
4364 assert( pPhrase->aToken[iToken].pSegcsr==0 );
4365 fts3EvalDlPhraseNext(pTab, &pPhrase->doclist, pbEof);
4366 p->pList = pPhrase->doclist.pList;
4367 p->nList = pPhrase->doclist.nList;
4368 p->iDocid = pPhrase->doclist.iDocid;
4369 }else{
4370 Fts3PhraseToken *pToken = &pPhrase->aToken[iToken];
4371 assert( pToken->pDeferred==0 );
4372 assert( pToken->pSegcsr || pPhrase->iDoclistToken>=0 );
4373 if( pToken->pSegcsr ){
4374 assert( p->bIgnore==0 );
4375 rc = sqlite3Fts3MsrIncrNext(
4376 pTab, pToken->pSegcsr, &p->iDocid, &p->pList, &p->nList
4377 );
4378 if( p->pList==0 ) *pbEof = 1;
4379 }else{
4380 p->bIgnore = 1;
4381 }
4382 }
4383
4384 return rc;
4385 }
4386
4387
4388 /*
4389 ** The phrase iterator passed as the second argument:
4390 **
4391 ** * features at least one token that uses an incremental doclist, and
4392 **
4393 ** * does not contain any deferred tokens.
4394 **
4395 ** Advance it to the next matching documnent in the database and populate
4396 ** the Fts3Doclist.pList and nList fields.
4397 **
4398 ** If there is no "next" entry and no error occurs, then *pbEof is set to
4399 ** 1 before returning. Otherwise, if no error occurs and the iterator is
4400 ** successfully advanced, *pbEof is set to 0.
4401 **
4402 ** If an error occurs, return an SQLite error code. Otherwise, return
4403 ** SQLITE_OK.
4404 */
4405 static int fts3EvalIncrPhraseNext(
4406 Fts3Cursor *pCsr, /* FTS Cursor handle */
4407 Fts3Phrase *p, /* Phrase object to advance to next docid */
4408 u8 *pbEof /* OUT: Set to 1 if EOF */
4409 ){
4410 int rc = SQLITE_OK;
4411 Fts3Doclist *pDL = &p->doclist;
4412 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
4413 u8 bEof = 0;
4414
4415 /* This is only called if it is guaranteed that the phrase has at least
4416 ** one incremental token. In which case the bIncr flag is set. */
4417 assert( p->bIncr==1 );
4418
4419 if( p->nToken==1 && p->bIncr ){
4420 rc = sqlite3Fts3MsrIncrNext(pTab, p->aToken[0].pSegcsr,
4421 &pDL->iDocid, &pDL->pList, &pDL->nList
4422 );
4423 if( pDL->pList==0 ) bEof = 1;
4424 }else{
4425 int bDescDoclist = pCsr->bDesc;
4426 struct TokenDoclist a[MAX_INCR_PHRASE_TOKENS];
4427
4428 memset(a, 0, sizeof(a));
4429 assert( p->nToken<=MAX_INCR_PHRASE_TOKENS );
4430 assert( p->iDoclistToken<MAX_INCR_PHRASE_TOKENS );
4431
4432 while( bEof==0 ){
4433 int bMaxSet = 0;
4434 sqlite3_int64 iMax = 0; /* Largest docid for all iterators */
4435 int i; /* Used to iterate through tokens */
4436
4437 /* Advance the iterator for each token in the phrase once. */
4438 for(i=0; rc==SQLITE_OK && i<p->nToken && bEof==0; i++){
4439 rc = incrPhraseTokenNext(pTab, p, i, &a[i], &bEof);
4440 if( a[i].bIgnore==0 && (bMaxSet==0 || DOCID_CMP(iMax, a[i].iDocid)<0) ){
4441 iMax = a[i].iDocid;
4442 bMaxSet = 1;
4443 }
4444 }
4445 assert( rc!=SQLITE_OK || (p->nToken>=1 && a[p->nToken-1].bIgnore==0) );
4446 assert( rc!=SQLITE_OK || bMaxSet );
4447
4448 /* Keep advancing iterators until they all point to the same document */
4449 for(i=0; i<p->nToken; i++){
4450 while( rc==SQLITE_OK && bEof==0
4451 && a[i].bIgnore==0 && DOCID_CMP(a[i].iDocid, iMax)<0
4452 ){
4453 rc = incrPhraseTokenNext(pTab, p, i, &a[i], &bEof);
4454 if( DOCID_CMP(a[i].iDocid, iMax)>0 ){
4455 iMax = a[i].iDocid;
4456 i = 0;
4457 }
4458 }
4459 }
4460
4461 /* Check if the current entries really are a phrase match */
4462 if( bEof==0 ){
4463 int nList = 0;
4464 int nByte = a[p->nToken-1].nList;
4465 char *aDoclist = sqlite3_malloc(nByte+1);
4466 if( !aDoclist ) return SQLITE_NOMEM;
4467 memcpy(aDoclist, a[p->nToken-1].pList, nByte+1);
4468
4469 for(i=0; i<(p->nToken-1); i++){
4470 if( a[i].bIgnore==0 ){
4471 char *pL = a[i].pList;
4472 char *pR = aDoclist;
4473 char *pOut = aDoclist;
4474 int nDist = p->nToken-1-i;
4475 int res = fts3PoslistPhraseMerge(&pOut, nDist, 0, 1, &pL, &pR);
4476 if( res==0 ) break;
4477 nList = (int)(pOut - aDoclist);
4478 }
4479 }
4480 if( i==(p->nToken-1) ){
4481 pDL->iDocid = iMax;
4482 pDL->pList = aDoclist;
4483 pDL->nList = nList;
4484 pDL->bFreeList = 1;
4485 break;
4486 }
4487 sqlite3_free(aDoclist);
4488 }
4489 }
4490 }
4491
4492 *pbEof = bEof;
4493 return rc;
4494 }
4495
4496 /*
4497 ** Attempt to move the phrase iterator to point to the next matching docid.
4498 ** If an error occurs, return an SQLite error code. Otherwise, return
4499 ** SQLITE_OK.
4500 **
4501 ** If there is no "next" entry and no error occurs, then *pbEof is set to
4502 ** 1 before returning. Otherwise, if no error occurs and the iterator is
4503 ** successfully advanced, *pbEof is set to 0.
4504 */
4505 static int fts3EvalPhraseNext(
4506 Fts3Cursor *pCsr, /* FTS Cursor handle */
4507 Fts3Phrase *p, /* Phrase object to advance to next docid */
4508 u8 *pbEof /* OUT: Set to 1 if EOF */
4509 ){
4510 int rc = SQLITE_OK;
4511 Fts3Doclist *pDL = &p->doclist;
4512 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
4513
4514 if( p->bIncr ){
4515 rc = fts3EvalIncrPhraseNext(pCsr, p, pbEof);
4516 }else if( pCsr->bDesc!=pTab->bDescIdx && pDL->nAll ){
4517 sqlite3Fts3DoclistPrev(pTab->bDescIdx, pDL->aAll, pDL->nAll,
4518 &pDL->pNextDocid, &pDL->iDocid, &pDL->nList, pbEof
4519 );
4520 pDL->pList = pDL->pNextDocid;
4521 }else{
4522 fts3EvalDlPhraseNext(pTab, pDL, pbEof);
4523 }
4524
4525 return rc;
4526 }
4527
4528 /*
4529 **
4530 ** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
4531 ** Otherwise, fts3EvalPhraseStart() is called on all phrases within the
4532 ** expression. Also the Fts3Expr.bDeferred variable is set to true for any
4533 ** expressions for which all descendent tokens are deferred.
4534 **
4535 ** If parameter bOptOk is zero, then it is guaranteed that the
4536 ** Fts3Phrase.doclist.aAll/nAll variables contain the entire doclist for
4537 ** each phrase in the expression (subject to deferred token processing).
4538 ** Or, if bOptOk is non-zero, then one or more tokens within the expression
4539 ** may be loaded incrementally, meaning doclist.aAll/nAll is not available.
4540 **
4541 ** If an error occurs within this function, *pRc is set to an SQLite error
4542 ** code before returning.
4543 */
4544 static void fts3EvalStartReaders(
4545 Fts3Cursor *pCsr, /* FTS Cursor handle */
4546 Fts3Expr *pExpr, /* Expression to initialize phrases in */
4547 int *pRc /* IN/OUT: Error code */
4548 ){
4549 if( pExpr && SQLITE_OK==*pRc ){
4550 if( pExpr->eType==FTSQUERY_PHRASE ){
4551 int i;
4552 int nToken = pExpr->pPhrase->nToken;
4553 for(i=0; i<nToken; i++){
4554 if( pExpr->pPhrase->aToken[i].pDeferred==0 ) break;
4555 }
4556 pExpr->bDeferred = (i==nToken);
4557 *pRc = fts3EvalPhraseStart(pCsr, 1, pExpr->pPhrase);
4558 }else{
4559 fts3EvalStartReaders(pCsr, pExpr->pLeft, pRc);
4560 fts3EvalStartReaders(pCsr, pExpr->pRight, pRc);
4561 pExpr->bDeferred = (pExpr->pLeft->bDeferred && pExpr->pRight->bDeferred);
4562 }
4563 }
4564 }
4565
4566 /*
4567 ** An array of the following structures is assembled as part of the process
4568 ** of selecting tokens to defer before the query starts executing (as part
4569 ** of the xFilter() method). There is one element in the array for each
4570 ** token in the FTS expression.
4571 **
4572 ** Tokens are divided into AND/NEAR clusters. All tokens in a cluster belong
4573 ** to phrases that are connected only by AND and NEAR operators (not OR or
4574 ** NOT). When determining tokens to defer, each AND/NEAR cluster is considered
4575 ** separately. The root of a tokens AND/NEAR cluster is stored in
4576 ** Fts3TokenAndCost.pRoot.
4577 */
4578 typedef struct Fts3TokenAndCost Fts3TokenAndCost;
4579 struct Fts3TokenAndCost {
4580 Fts3Phrase *pPhrase; /* The phrase the token belongs to */
4581 int iToken; /* Position of token in phrase */
4582 Fts3PhraseToken *pToken; /* The token itself */
4583 Fts3Expr *pRoot; /* Root of NEAR/AND cluster */
4584 int nOvfl; /* Number of overflow pages to load doclist */
4585 int iCol; /* The column the token must match */
4586 };
4587
4588 /*
4589 ** This function is used to populate an allocated Fts3TokenAndCost array.
4590 **
4591 ** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
4592 ** Otherwise, if an error occurs during execution, *pRc is set to an
4593 ** SQLite error code.
4594 */
4595 static void fts3EvalTokenCosts(
4596 Fts3Cursor *pCsr, /* FTS Cursor handle */
4597 Fts3Expr *pRoot, /* Root of current AND/NEAR cluster */
4598 Fts3Expr *pExpr, /* Expression to consider */
4599 Fts3TokenAndCost **ppTC, /* Write new entries to *(*ppTC)++ */
4600 Fts3Expr ***ppOr, /* Write new OR root to *(*ppOr)++ */
4601 int *pRc /* IN/OUT: Error code */
4602 ){
4603 if( *pRc==SQLITE_OK ){
4604 if( pExpr->eType==FTSQUERY_PHRASE ){
4605 Fts3Phrase *pPhrase = pExpr->pPhrase;
4606 int i;
4607 for(i=0; *pRc==SQLITE_OK && i<pPhrase->nToken; i++){
4608 Fts3TokenAndCost *pTC = (*ppTC)++;
4609 pTC->pPhrase = pPhrase;
4610 pTC->iToken = i;
4611 pTC->pRoot = pRoot;
4612 pTC->pToken = &pPhrase->aToken[i];
4613 pTC->iCol = pPhrase->iColumn;
4614 *pRc = sqlite3Fts3MsrOvfl(pCsr, pTC->pToken->pSegcsr, &pTC->nOvfl);
4615 }
4616 }else if( pExpr->eType!=FTSQUERY_NOT ){
4617 assert( pExpr->eType==FTSQUERY_OR
4618 || pExpr->eType==FTSQUERY_AND
4619 || pExpr->eType==FTSQUERY_NEAR
4620 );
4621 assert( pExpr->pLeft && pExpr->pRight );
4622 if( pExpr->eType==FTSQUERY_OR ){
4623 pRoot = pExpr->pLeft;
4624 **ppOr = pRoot;
4625 (*ppOr)++;
4626 }
4627 fts3EvalTokenCosts(pCsr, pRoot, pExpr->pLeft, ppTC, ppOr, pRc);
4628 if( pExpr->eType==FTSQUERY_OR ){
4629 pRoot = pExpr->pRight;
4630 **ppOr = pRoot;
4631 (*ppOr)++;
4632 }
4633 fts3EvalTokenCosts(pCsr, pRoot, pExpr->pRight, ppTC, ppOr, pRc);
4634 }
4635 }
4636 }
4637
4638 /*
4639 ** Determine the average document (row) size in pages. If successful,
4640 ** write this value to *pnPage and return SQLITE_OK. Otherwise, return
4641 ** an SQLite error code.
4642 **
4643 ** The average document size in pages is calculated by first calculating
4644 ** determining the average size in bytes, B. If B is less than the amount
4645 ** of data that will fit on a single leaf page of an intkey table in
4646 ** this database, then the average docsize is 1. Otherwise, it is 1 plus
4647 ** the number of overflow pages consumed by a record B bytes in size.
4648 */
4649 static int fts3EvalAverageDocsize(Fts3Cursor *pCsr, int *pnPage){
4650 if( pCsr->nRowAvg==0 ){
4651 /* The average document size, which is required to calculate the cost
4652 ** of each doclist, has not yet been determined. Read the required
4653 ** data from the %_stat table to calculate it.
4654 **
4655 ** Entry 0 of the %_stat table is a blob containing (nCol+1) FTS3
4656 ** varints, where nCol is the number of columns in the FTS3 table.
4657 ** The first varint is the number of documents currently stored in
4658 ** the table. The following nCol varints contain the total amount of
4659 ** data stored in all rows of each column of the table, from left
4660 ** to right.
4661 */
4662 int rc;
4663 Fts3Table *p = (Fts3Table*)pCsr->base.pVtab;
4664 sqlite3_stmt *pStmt;
4665 sqlite3_int64 nDoc = 0;
4666 sqlite3_int64 nByte = 0;
4667 const char *pEnd;
4668 const char *a;
4669
4670 rc = sqlite3Fts3SelectDoctotal(p, &pStmt);
4671 if( rc!=SQLITE_OK ) return rc;
4672 a = sqlite3_column_blob(pStmt, 0);
4673 assert( a );
4674
4675 pEnd = &a[sqlite3_column_bytes(pStmt, 0)];
4676 a += sqlite3Fts3GetVarint(a, &nDoc);
4677 while( a<pEnd ){
4678 a += sqlite3Fts3GetVarint(a, &nByte);
4679 }
4680 if( nDoc==0 || nByte==0 ){
4681 sqlite3_reset(pStmt);
4682 return FTS_CORRUPT_VTAB;
4683 }
4684
4685 pCsr->nDoc = nDoc;
4686 pCsr->nRowAvg = (int)(((nByte / nDoc) + p->nPgsz) / p->nPgsz);
4687 assert( pCsr->nRowAvg>0 );
4688 rc = sqlite3_reset(pStmt);
4689 if( rc!=SQLITE_OK ) return rc;
4690 }
4691
4692 *pnPage = pCsr->nRowAvg;
4693 return SQLITE_OK;
4694 }
4695
4696 /*
4697 ** This function is called to select the tokens (if any) that will be
4698 ** deferred. The array aTC[] has already been populated when this is
4699 ** called.
4700 **
4701 ** This function is called once for each AND/NEAR cluster in the
4702 ** expression. Each invocation determines which tokens to defer within
4703 ** the cluster with root node pRoot. See comments above the definition
4704 ** of struct Fts3TokenAndCost for more details.
4705 **
4706 ** If no error occurs, SQLITE_OK is returned and sqlite3Fts3DeferToken()
4707 ** called on each token to defer. Otherwise, an SQLite error code is
4708 ** returned.
4709 */
4710 static int fts3EvalSelectDeferred(
4711 Fts3Cursor *pCsr, /* FTS Cursor handle */
4712 Fts3Expr *pRoot, /* Consider tokens with this root node */
4713 Fts3TokenAndCost *aTC, /* Array of expression tokens and costs */
4714 int nTC /* Number of entries in aTC[] */
4715 ){
4716 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
4717 int nDocSize = 0; /* Number of pages per doc loaded */
4718 int rc = SQLITE_OK; /* Return code */
4719 int ii; /* Iterator variable for various purposes */
4720 int nOvfl = 0; /* Total overflow pages used by doclists */
4721 int nToken = 0; /* Total number of tokens in cluster */
4722
4723 int nMinEst = 0; /* The minimum count for any phrase so far. */
4724 int nLoad4 = 1; /* (Phrases that will be loaded)^4. */
4725
4726 /* Tokens are never deferred for FTS tables created using the content=xxx
4727 ** option. The reason being that it is not guaranteed that the content
4728 ** table actually contains the same data as the index. To prevent this from
4729 ** causing any problems, the deferred token optimization is completely
4730 ** disabled for content=xxx tables. */
4731 if( pTab->zContentTbl ){
4732 return SQLITE_OK;
4733 }
4734
4735 /* Count the tokens in this AND/NEAR cluster. If none of the doclists
4736 ** associated with the tokens spill onto overflow pages, or if there is
4737 ** only 1 token, exit early. No tokens to defer in this case. */
4738 for(ii=0; ii<nTC; ii++){
4739 if( aTC[ii].pRoot==pRoot ){
4740 nOvfl += aTC[ii].nOvfl;
4741 nToken++;
4742 }
4743 }
4744 if( nOvfl==0 || nToken<2 ) return SQLITE_OK;
4745
4746 /* Obtain the average docsize (in pages). */
4747 rc = fts3EvalAverageDocsize(pCsr, &nDocSize);
4748 assert( rc!=SQLITE_OK || nDocSize>0 );
4749
4750
4751 /* Iterate through all tokens in this AND/NEAR cluster, in ascending order
4752 ** of the number of overflow pages that will be loaded by the pager layer
4753 ** to retrieve the entire doclist for the token from the full-text index.
4754 ** Load the doclists for tokens that are either:
4755 **
4756 ** a. The cheapest token in the entire query (i.e. the one visited by the
4757 ** first iteration of this loop), or
4758 **
4759 ** b. Part of a multi-token phrase.
4760 **
4761 ** After each token doclist is loaded, merge it with the others from the
4762 ** same phrase and count the number of documents that the merged doclist
4763 ** contains. Set variable "nMinEst" to the smallest number of documents in
4764 ** any phrase doclist for which 1 or more token doclists have been loaded.
4765 ** Let nOther be the number of other phrases for which it is certain that
4766 ** one or more tokens will not be deferred.
4767 **
4768 ** Then, for each token, defer it if loading the doclist would result in
4769 ** loading N or more overflow pages into memory, where N is computed as:
4770 **
4771 ** (nMinEst + 4^nOther - 1) / (4^nOther)
4772 */
4773 for(ii=0; ii<nToken && rc==SQLITE_OK; ii++){
4774 int iTC; /* Used to iterate through aTC[] array. */
4775 Fts3TokenAndCost *pTC = 0; /* Set to cheapest remaining token. */
4776
4777 /* Set pTC to point to the cheapest remaining token. */
4778 for(iTC=0; iTC<nTC; iTC++){
4779 if( aTC[iTC].pToken && aTC[iTC].pRoot==pRoot
4780 && (!pTC || aTC[iTC].nOvfl<pTC->nOvfl)
4781 ){
4782 pTC = &aTC[iTC];
4783 }
4784 }
4785 assert( pTC );
4786
4787 if( ii && pTC->nOvfl>=((nMinEst+(nLoad4/4)-1)/(nLoad4/4))*nDocSize ){
4788 /* The number of overflow pages to load for this (and therefore all
4789 ** subsequent) tokens is greater than the estimated number of pages
4790 ** that will be loaded if all subsequent tokens are deferred.
4791 */
4792 Fts3PhraseToken *pToken = pTC->pToken;
4793 rc = sqlite3Fts3DeferToken(pCsr, pToken, pTC->iCol);
4794 fts3SegReaderCursorFree(pToken->pSegcsr);
4795 pToken->pSegcsr = 0;
4796 }else{
4797 /* Set nLoad4 to the value of (4^nOther) for the next iteration of the
4798 ** for-loop. Except, limit the value to 2^24 to prevent it from
4799 ** overflowing the 32-bit integer it is stored in. */
4800 if( ii<12 ) nLoad4 = nLoad4*4;
4801
4802 if( ii==0 || (pTC->pPhrase->nToken>1 && ii!=nToken-1) ){
4803 /* Either this is the cheapest token in the entire query, or it is
4804 ** part of a multi-token phrase. Either way, the entire doclist will
4805 ** (eventually) be loaded into memory. It may as well be now. */
4806 Fts3PhraseToken *pToken = pTC->pToken;
4807 int nList = 0;
4808 char *pList = 0;
4809 rc = fts3TermSelect(pTab, pToken, pTC->iCol, &nList, &pList);
4810 assert( rc==SQLITE_OK || pList==0 );
4811 if( rc==SQLITE_OK ){
4812 int nCount;
4813 fts3EvalPhraseMergeToken(pTab, pTC->pPhrase, pTC->iToken,pList,nList);
4814 nCount = fts3DoclistCountDocids(
4815 pTC->pPhrase->doclist.aAll, pTC->pPhrase->doclist.nAll
4816 );
4817 if( ii==0 || nCount<nMinEst ) nMinEst = nCount;
4818 }
4819 }
4820 }
4821 pTC->pToken = 0;
4822 }
4823
4824 return rc;
4825 }
4826
4827 /*
4828 ** This function is called from within the xFilter method. It initializes
4829 ** the full-text query currently stored in pCsr->pExpr. To iterate through
4830 ** the results of a query, the caller does:
4831 **
4832 ** fts3EvalStart(pCsr);
4833 ** while( 1 ){
4834 ** fts3EvalNext(pCsr);
4835 ** if( pCsr->bEof ) break;
4836 ** ... return row pCsr->iPrevId to the caller ...
4837 ** }
4838 */
4839 static int fts3EvalStart(Fts3Cursor *pCsr){
4840 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
4841 int rc = SQLITE_OK;
4842 int nToken = 0;
4843 int nOr = 0;
4844
4845 /* Allocate a MultiSegReader for each token in the expression. */
4846 fts3EvalAllocateReaders(pCsr, pCsr->pExpr, &nToken, &nOr, &rc);
4847
4848 /* Determine which, if any, tokens in the expression should be deferred. */
4849 #ifndef SQLITE_DISABLE_FTS4_DEFERRED
4850 if( rc==SQLITE_OK && nToken>1 && pTab->bFts4 ){
4851 Fts3TokenAndCost *aTC;
4852 Fts3Expr **apOr;
4853 aTC = (Fts3TokenAndCost *)sqlite3_malloc(
4854 sizeof(Fts3TokenAndCost) * nToken
4855 + sizeof(Fts3Expr *) * nOr * 2
4856 );
4857 apOr = (Fts3Expr **)&aTC[nToken];
4858
4859 if( !aTC ){
4860 rc = SQLITE_NOMEM;
4861 }else{
4862 int ii;
4863 Fts3TokenAndCost *pTC = aTC;
4864 Fts3Expr **ppOr = apOr;
4865
4866 fts3EvalTokenCosts(pCsr, 0, pCsr->pExpr, &pTC, &ppOr, &rc);
4867 nToken = (int)(pTC-aTC);
4868 nOr = (int)(ppOr-apOr);
4869
4870 if( rc==SQLITE_OK ){
4871 rc = fts3EvalSelectDeferred(pCsr, 0, aTC, nToken);
4872 for(ii=0; rc==SQLITE_OK && ii<nOr; ii++){
4873 rc = fts3EvalSelectDeferred(pCsr, apOr[ii], aTC, nToken);
4874 }
4875 }
4876
4877 sqlite3_free(aTC);
4878 }
4879 }
4880 #endif
4881
4882 fts3EvalStartReaders(pCsr, pCsr->pExpr, &rc);
4883 return rc;
4884 }
4885
4886 /*
4887 ** Invalidate the current position list for phrase pPhrase.
4888 */
4889 static void fts3EvalInvalidatePoslist(Fts3Phrase *pPhrase){
4890 if( pPhrase->doclist.bFreeList ){
4891 sqlite3_free(pPhrase->doclist.pList);
4892 }
4893 pPhrase->doclist.pList = 0;
4894 pPhrase->doclist.nList = 0;
4895 pPhrase->doclist.bFreeList = 0;
4896 }
4897
4898 /*
4899 ** This function is called to edit the position list associated with
4900 ** the phrase object passed as the fifth argument according to a NEAR
4901 ** condition. For example:
4902 **
4903 ** abc NEAR/5 "def ghi"
4904 **
4905 ** Parameter nNear is passed the NEAR distance of the expression (5 in
4906 ** the example above). When this function is called, *paPoslist points to
4907 ** the position list, and *pnToken is the number of phrase tokens in, the
4908 ** phrase on the other side of the NEAR operator to pPhrase. For example,
4909 ** if pPhrase refers to the "def ghi" phrase, then *paPoslist points to
4910 ** the position list associated with phrase "abc".
4911 **
4912 ** All positions in the pPhrase position list that are not sufficiently
4913 ** close to a position in the *paPoslist position list are removed. If this
4914 ** leaves 0 positions, zero is returned. Otherwise, non-zero.
4915 **
4916 ** Before returning, *paPoslist is set to point to the position lsit
4917 ** associated with pPhrase. And *pnToken is set to the number of tokens in
4918 ** pPhrase.
4919 */
4920 static int fts3EvalNearTrim(
4921 int nNear, /* NEAR distance. As in "NEAR/nNear". */
4922 char *aTmp, /* Temporary space to use */
4923 char **paPoslist, /* IN/OUT: Position list */
4924 int *pnToken, /* IN/OUT: Tokens in phrase of *paPoslist */
4925 Fts3Phrase *pPhrase /* The phrase object to trim the doclist of */
4926 ){
4927 int nParam1 = nNear + pPhrase->nToken;
4928 int nParam2 = nNear + *pnToken;
4929 int nNew;
4930 char *p2;
4931 char *pOut;
4932 int res;
4933
4934 assert( pPhrase->doclist.pList );
4935
4936 p2 = pOut = pPhrase->doclist.pList;
4937 res = fts3PoslistNearMerge(
4938 &pOut, aTmp, nParam1, nParam2, paPoslist, &p2
4939 );
4940 if( res ){
4941 nNew = (int)(pOut - pPhrase->doclist.pList) - 1;
4942 assert( pPhrase->doclist.pList[nNew]=='\0' );
4943 assert( nNew<=pPhrase->doclist.nList && nNew>0 );
4944 memset(&pPhrase->doclist.pList[nNew], 0, pPhrase->doclist.nList - nNew);
4945 pPhrase->doclist.nList = nNew;
4946 *paPoslist = pPhrase->doclist.pList;
4947 *pnToken = pPhrase->nToken;
4948 }
4949
4950 return res;
4951 }
4952
4953 /*
4954 ** This function is a no-op if *pRc is other than SQLITE_OK when it is called.
4955 ** Otherwise, it advances the expression passed as the second argument to
4956 ** point to the next matching row in the database. Expressions iterate through
4957 ** matching rows in docid order. Ascending order if Fts3Cursor.bDesc is zero,
4958 ** or descending if it is non-zero.
4959 **
4960 ** If an error occurs, *pRc is set to an SQLite error code. Otherwise, if
4961 ** successful, the following variables in pExpr are set:
4962 **
4963 ** Fts3Expr.bEof (non-zero if EOF - there is no next row)
4964 ** Fts3Expr.iDocid (valid if bEof==0. The docid of the next row)
4965 **
4966 ** If the expression is of type FTSQUERY_PHRASE, and the expression is not
4967 ** at EOF, then the following variables are populated with the position list
4968 ** for the phrase for the visited row:
4969 **
4970 ** FTs3Expr.pPhrase->doclist.nList (length of pList in bytes)
4971 ** FTs3Expr.pPhrase->doclist.pList (pointer to position list)
4972 **
4973 ** It says above that this function advances the expression to the next
4974 ** matching row. This is usually true, but there are the following exceptions:
4975 **
4976 ** 1. Deferred tokens are not taken into account. If a phrase consists
4977 ** entirely of deferred tokens, it is assumed to match every row in
4978 ** the db. In this case the position-list is not populated at all.
4979 **
4980 ** Or, if a phrase contains one or more deferred tokens and one or
4981 ** more non-deferred tokens, then the expression is advanced to the
4982 ** next possible match, considering only non-deferred tokens. In other
4983 ** words, if the phrase is "A B C", and "B" is deferred, the expression
4984 ** is advanced to the next row that contains an instance of "A * C",
4985 ** where "*" may match any single token. The position list in this case
4986 ** is populated as for "A * C" before returning.
4987 **
4988 ** 2. NEAR is treated as AND. If the expression is "x NEAR y", it is
4989 ** advanced to point to the next row that matches "x AND y".
4990 **
4991 ** See fts3EvalTestDeferredAndNear() for details on testing if a row is
4992 ** really a match, taking into account deferred tokens and NEAR operators.
4993 */
4994 static void fts3EvalNextRow(
4995 Fts3Cursor *pCsr, /* FTS Cursor handle */
4996 Fts3Expr *pExpr, /* Expr. to advance to next matching row */
4997 int *pRc /* IN/OUT: Error code */
4998 ){
4999 if( *pRc==SQLITE_OK ){
5000 int bDescDoclist = pCsr->bDesc; /* Used by DOCID_CMP() macro */
5001 assert( pExpr->bEof==0 );
5002 pExpr->bStart = 1;
5003
5004 switch( pExpr->eType ){
5005 case FTSQUERY_NEAR:
5006 case FTSQUERY_AND: {
5007 Fts3Expr *pLeft = pExpr->pLeft;
5008 Fts3Expr *pRight = pExpr->pRight;
5009 assert( !pLeft->bDeferred || !pRight->bDeferred );
5010
5011 if( pLeft->bDeferred ){
5012 /* LHS is entirely deferred. So we assume it matches every row.
5013 ** Advance the RHS iterator to find the next row visited. */
5014 fts3EvalNextRow(pCsr, pRight, pRc);
5015 pExpr->iDocid = pRight->iDocid;
5016 pExpr->bEof = pRight->bEof;
5017 }else if( pRight->bDeferred ){
5018 /* RHS is entirely deferred. So we assume it matches every row.
5019 ** Advance the LHS iterator to find the next row visited. */
5020 fts3EvalNextRow(pCsr, pLeft, pRc);
5021 pExpr->iDocid = pLeft->iDocid;
5022 pExpr->bEof = pLeft->bEof;
5023 }else{
5024 /* Neither the RHS or LHS are deferred. */
5025 fts3EvalNextRow(pCsr, pLeft, pRc);
5026 fts3EvalNextRow(pCsr, pRight, pRc);
5027 while( !pLeft->bEof && !pRight->bEof && *pRc==SQLITE_OK ){
5028 sqlite3_int64 iDiff = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
5029 if( iDiff==0 ) break;
5030 if( iDiff<0 ){
5031 fts3EvalNextRow(pCsr, pLeft, pRc);
5032 }else{
5033 fts3EvalNextRow(pCsr, pRight, pRc);
5034 }
5035 }
5036 pExpr->iDocid = pLeft->iDocid;
5037 pExpr->bEof = (pLeft->bEof || pRight->bEof);
5038 }
5039 break;
5040 }
5041
5042 case FTSQUERY_OR: {
5043 Fts3Expr *pLeft = pExpr->pLeft;
5044 Fts3Expr *pRight = pExpr->pRight;
5045 sqlite3_int64 iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
5046
5047 assert( pLeft->bStart || pLeft->iDocid==pRight->iDocid );
5048 assert( pRight->bStart || pLeft->iDocid==pRight->iDocid );
5049
5050 if( pRight->bEof || (pLeft->bEof==0 && iCmp<0) ){
5051 fts3EvalNextRow(pCsr, pLeft, pRc);
5052 }else if( pLeft->bEof || (pRight->bEof==0 && iCmp>0) ){
5053 fts3EvalNextRow(pCsr, pRight, pRc);
5054 }else{
5055 fts3EvalNextRow(pCsr, pLeft, pRc);
5056 fts3EvalNextRow(pCsr, pRight, pRc);
5057 }
5058
5059 pExpr->bEof = (pLeft->bEof && pRight->bEof);
5060 iCmp = DOCID_CMP(pLeft->iDocid, pRight->iDocid);
5061 if( pRight->bEof || (pLeft->bEof==0 && iCmp<0) ){
5062 pExpr->iDocid = pLeft->iDocid;
5063 }else{
5064 pExpr->iDocid = pRight->iDocid;
5065 }
5066
5067 break;
5068 }
5069
5070 case FTSQUERY_NOT: {
5071 Fts3Expr *pLeft = pExpr->pLeft;
5072 Fts3Expr *pRight = pExpr->pRight;
5073
5074 if( pRight->bStart==0 ){
5075 fts3EvalNextRow(pCsr, pRight, pRc);
5076 assert( *pRc!=SQLITE_OK || pRight->bStart );
5077 }
5078
5079 fts3EvalNextRow(pCsr, pLeft, pRc);
5080 if( pLeft->bEof==0 ){
5081 while( !*pRc
5082 && !pRight->bEof
5083 && DOCID_CMP(pLeft->iDocid, pRight->iDocid)>0
5084 ){
5085 fts3EvalNextRow(pCsr, pRight, pRc);
5086 }
5087 }
5088 pExpr->iDocid = pLeft->iDocid;
5089 pExpr->bEof = pLeft->bEof;
5090 break;
5091 }
5092
5093 default: {
5094 Fts3Phrase *pPhrase = pExpr->pPhrase;
5095 fts3EvalInvalidatePoslist(pPhrase);
5096 *pRc = fts3EvalPhraseNext(pCsr, pPhrase, &pExpr->bEof);
5097 pExpr->iDocid = pPhrase->doclist.iDocid;
5098 break;
5099 }
5100 }
5101 }
5102 }
5103
5104 /*
5105 ** If *pRc is not SQLITE_OK, or if pExpr is not the root node of a NEAR
5106 ** cluster, then this function returns 1 immediately.
5107 **
5108 ** Otherwise, it checks if the current row really does match the NEAR
5109 ** expression, using the data currently stored in the position lists
5110 ** (Fts3Expr->pPhrase.doclist.pList/nList) for each phrase in the expression.
5111 **
5112 ** If the current row is a match, the position list associated with each
5113 ** phrase in the NEAR expression is edited in place to contain only those
5114 ** phrase instances sufficiently close to their peers to satisfy all NEAR
5115 ** constraints. In this case it returns 1. If the NEAR expression does not
5116 ** match the current row, 0 is returned. The position lists may or may not
5117 ** be edited if 0 is returned.
5118 */
5119 static int fts3EvalNearTest(Fts3Expr *pExpr, int *pRc){
5120 int res = 1;
5121
5122 /* The following block runs if pExpr is the root of a NEAR query.
5123 ** For example, the query:
5124 **
5125 ** "w" NEAR "x" NEAR "y" NEAR "z"
5126 **
5127 ** which is represented in tree form as:
5128 **
5129 ** |
5130 ** +--NEAR--+ <-- root of NEAR query
5131 ** | |
5132 ** +--NEAR--+ "z"
5133 ** | |
5134 ** +--NEAR--+ "y"
5135 ** | |
5136 ** "w" "x"
5137 **
5138 ** The right-hand child of a NEAR node is always a phrase. The
5139 ** left-hand child may be either a phrase or a NEAR node. There are
5140 ** no exceptions to this - it's the way the parser in fts3_expr.c works.
5141 */
5142 if( *pRc==SQLITE_OK
5143 && pExpr->eType==FTSQUERY_NEAR
5144 && pExpr->bEof==0
5145 && (pExpr->pParent==0 || pExpr->pParent->eType!=FTSQUERY_NEAR)
5146 ){
5147 Fts3Expr *p;
5148 int nTmp = 0; /* Bytes of temp space */
5149 char *aTmp; /* Temp space for PoslistNearMerge() */
5150
5151 /* Allocate temporary working space. */
5152 for(p=pExpr; p->pLeft; p=p->pLeft){
5153 nTmp += p->pRight->pPhrase->doclist.nList;
5154 }
5155 nTmp += p->pPhrase->doclist.nList;
5156 if( nTmp==0 ){
5157 res = 0;
5158 }else{
5159 aTmp = sqlite3_malloc(nTmp*2);
5160 if( !aTmp ){
5161 *pRc = SQLITE_NOMEM;
5162 res = 0;
5163 }else{
5164 char *aPoslist = p->pPhrase->doclist.pList;
5165 int nToken = p->pPhrase->nToken;
5166
5167 for(p=p->pParent;res && p && p->eType==FTSQUERY_NEAR; p=p->pParent){
5168 Fts3Phrase *pPhrase = p->pRight->pPhrase;
5169 int nNear = p->nNear;
5170 res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase);
5171 }
5172
5173 aPoslist = pExpr->pRight->pPhrase->doclist.pList;
5174 nToken = pExpr->pRight->pPhrase->nToken;
5175 for(p=pExpr->pLeft; p && res; p=p->pLeft){
5176 int nNear;
5177 Fts3Phrase *pPhrase;
5178 assert( p->pParent && p->pParent->pLeft==p );
5179 nNear = p->pParent->nNear;
5180 pPhrase = (
5181 p->eType==FTSQUERY_NEAR ? p->pRight->pPhrase : p->pPhrase
5182 );
5183 res = fts3EvalNearTrim(nNear, aTmp, &aPoslist, &nToken, pPhrase);
5184 }
5185 }
5186
5187 sqlite3_free(aTmp);
5188 }
5189 }
5190
5191 return res;
5192 }
5193
5194 /*
5195 ** This function is a helper function for fts3EvalTestDeferredAndNear().
5196 ** Assuming no error occurs or has occurred, It returns non-zero if the
5197 ** expression passed as the second argument matches the row that pCsr
5198 ** currently points to, or zero if it does not.
5199 **
5200 ** If *pRc is not SQLITE_OK when this function is called, it is a no-op.
5201 ** If an error occurs during execution of this function, *pRc is set to
5202 ** the appropriate SQLite error code. In this case the returned value is
5203 ** undefined.
5204 */
5205 static int fts3EvalTestExpr(
5206 Fts3Cursor *pCsr, /* FTS cursor handle */
5207 Fts3Expr *pExpr, /* Expr to test. May or may not be root. */
5208 int *pRc /* IN/OUT: Error code */
5209 ){
5210 int bHit = 1; /* Return value */
5211 if( *pRc==SQLITE_OK ){
5212 switch( pExpr->eType ){
5213 case FTSQUERY_NEAR:
5214 case FTSQUERY_AND:
5215 bHit = (
5216 fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc)
5217 && fts3EvalTestExpr(pCsr, pExpr->pRight, pRc)
5218 && fts3EvalNearTest(pExpr, pRc)
5219 );
5220
5221 /* If the NEAR expression does not match any rows, zero the doclist for
5222 ** all phrases involved in the NEAR. This is because the snippet(),
5223 ** offsets() and matchinfo() functions are not supposed to recognize
5224 ** any instances of phrases that are part of unmatched NEAR queries.
5225 ** For example if this expression:
5226 **
5227 ** ... MATCH 'a OR (b NEAR c)'
5228 **
5229 ** is matched against a row containing:
5230 **
5231 ** 'a b d e'
5232 **
5233 ** then any snippet() should ony highlight the "a" term, not the "b"
5234 ** (as "b" is part of a non-matching NEAR clause).
5235 */
5236 if( bHit==0
5237 && pExpr->eType==FTSQUERY_NEAR
5238 && (pExpr->pParent==0 || pExpr->pParent->eType!=FTSQUERY_NEAR)
5239 ){
5240 Fts3Expr *p;
5241 for(p=pExpr; p->pPhrase==0; p=p->pLeft){
5242 if( p->pRight->iDocid==pCsr->iPrevId ){
5243 fts3EvalInvalidatePoslist(p->pRight->pPhrase);
5244 }
5245 }
5246 if( p->iDocid==pCsr->iPrevId ){
5247 fts3EvalInvalidatePoslist(p->pPhrase);
5248 }
5249 }
5250
5251 break;
5252
5253 case FTSQUERY_OR: {
5254 int bHit1 = fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc);
5255 int bHit2 = fts3EvalTestExpr(pCsr, pExpr->pRight, pRc);
5256 bHit = bHit1 || bHit2;
5257 break;
5258 }
5259
5260 case FTSQUERY_NOT:
5261 bHit = (
5262 fts3EvalTestExpr(pCsr, pExpr->pLeft, pRc)
5263 && !fts3EvalTestExpr(pCsr, pExpr->pRight, pRc)
5264 );
5265 break;
5266
5267 default: {
5268 #ifndef SQLITE_DISABLE_FTS4_DEFERRED
5269 if( pCsr->pDeferred
5270 && (pExpr->iDocid==pCsr->iPrevId || pExpr->bDeferred)
5271 ){
5272 Fts3Phrase *pPhrase = pExpr->pPhrase;
5273 assert( pExpr->bDeferred || pPhrase->doclist.bFreeList==0 );
5274 if( pExpr->bDeferred ){
5275 fts3EvalInvalidatePoslist(pPhrase);
5276 }
5277 *pRc = fts3EvalDeferredPhrase(pCsr, pPhrase);
5278 bHit = (pPhrase->doclist.pList!=0);
5279 pExpr->iDocid = pCsr->iPrevId;
5280 }else
5281 #endif
5282 {
5283 bHit = (pExpr->bEof==0 && pExpr->iDocid==pCsr->iPrevId);
5284 }
5285 break;
5286 }
5287 }
5288 }
5289 return bHit;
5290 }
5291
5292 /*
5293 ** This function is called as the second part of each xNext operation when
5294 ** iterating through the results of a full-text query. At this point the
5295 ** cursor points to a row that matches the query expression, with the
5296 ** following caveats:
5297 **
5298 ** * Up until this point, "NEAR" operators in the expression have been
5299 ** treated as "AND".
5300 **
5301 ** * Deferred tokens have not yet been considered.
5302 **
5303 ** If *pRc is not SQLITE_OK when this function is called, it immediately
5304 ** returns 0. Otherwise, it tests whether or not after considering NEAR
5305 ** operators and deferred tokens the current row is still a match for the
5306 ** expression. It returns 1 if both of the following are true:
5307 **
5308 ** 1. *pRc is SQLITE_OK when this function returns, and
5309 **
5310 ** 2. After scanning the current FTS table row for the deferred tokens,
5311 ** it is determined that the row does *not* match the query.
5312 **
5313 ** Or, if no error occurs and it seems the current row does match the FTS
5314 ** query, return 0.
5315 */
5316 static int fts3EvalTestDeferredAndNear(Fts3Cursor *pCsr, int *pRc){
5317 int rc = *pRc;
5318 int bMiss = 0;
5319 if( rc==SQLITE_OK ){
5320
5321 /* If there are one or more deferred tokens, load the current row into
5322 ** memory and scan it to determine the position list for each deferred
5323 ** token. Then, see if this row is really a match, considering deferred
5324 ** tokens and NEAR operators (neither of which were taken into account
5325 ** earlier, by fts3EvalNextRow()).
5326 */
5327 if( pCsr->pDeferred ){
5328 rc = fts3CursorSeek(0, pCsr);
5329 if( rc==SQLITE_OK ){
5330 rc = sqlite3Fts3CacheDeferredDoclists(pCsr);
5331 }
5332 }
5333 bMiss = (0==fts3EvalTestExpr(pCsr, pCsr->pExpr, &rc));
5334
5335 /* Free the position-lists accumulated for each deferred token above. */
5336 sqlite3Fts3FreeDeferredDoclists(pCsr);
5337 *pRc = rc;
5338 }
5339 return (rc==SQLITE_OK && bMiss);
5340 }
5341
5342 /*
5343 ** Advance to the next document that matches the FTS expression in
5344 ** Fts3Cursor.pExpr.
5345 */
5346 static int fts3EvalNext(Fts3Cursor *pCsr){
5347 int rc = SQLITE_OK; /* Return Code */
5348 Fts3Expr *pExpr = pCsr->pExpr;
5349 assert( pCsr->isEof==0 );
5350 if( pExpr==0 ){
5351 pCsr->isEof = 1;
5352 }else{
5353 do {
5354 if( pCsr->isRequireSeek==0 ){
5355 sqlite3_reset(pCsr->pStmt);
5356 }
5357 assert( sqlite3_data_count(pCsr->pStmt)==0 );
5358 fts3EvalNextRow(pCsr, pExpr, &rc);
5359 pCsr->isEof = pExpr->bEof;
5360 pCsr->isRequireSeek = 1;
5361 pCsr->isMatchinfoNeeded = 1;
5362 pCsr->iPrevId = pExpr->iDocid;
5363 }while( pCsr->isEof==0 && fts3EvalTestDeferredAndNear(pCsr, &rc) );
5364 }
5365
5366 /* Check if the cursor is past the end of the docid range specified
5367 ** by Fts3Cursor.iMinDocid/iMaxDocid. If so, set the EOF flag. */
5368 if( rc==SQLITE_OK && (
5369 (pCsr->bDesc==0 && pCsr->iPrevId>pCsr->iMaxDocid)
5370 || (pCsr->bDesc!=0 && pCsr->iPrevId<pCsr->iMinDocid)
5371 )){
5372 pCsr->isEof = 1;
5373 }
5374
5375 return rc;
5376 }
5377
5378 /*
5379 ** Restart interation for expression pExpr so that the next call to
5380 ** fts3EvalNext() visits the first row. Do not allow incremental
5381 ** loading or merging of phrase doclists for this iteration.
5382 **
5383 ** If *pRc is other than SQLITE_OK when this function is called, it is
5384 ** a no-op. If an error occurs within this function, *pRc is set to an
5385 ** SQLite error code before returning.
5386 */
5387 static void fts3EvalRestart(
5388 Fts3Cursor *pCsr,
5389 Fts3Expr *pExpr,
5390 int *pRc
5391 ){
5392 if( pExpr && *pRc==SQLITE_OK ){
5393 Fts3Phrase *pPhrase = pExpr->pPhrase;
5394
5395 if( pPhrase ){
5396 fts3EvalInvalidatePoslist(pPhrase);
5397 if( pPhrase->bIncr ){
5398 int i;
5399 for(i=0; i<pPhrase->nToken; i++){
5400 Fts3PhraseToken *pToken = &pPhrase->aToken[i];
5401 assert( pToken->pDeferred==0 );
5402 if( pToken->pSegcsr ){
5403 sqlite3Fts3MsrIncrRestart(pToken->pSegcsr);
5404 }
5405 }
5406 *pRc = fts3EvalPhraseStart(pCsr, 0, pPhrase);
5407 }
5408 pPhrase->doclist.pNextDocid = 0;
5409 pPhrase->doclist.iDocid = 0;
5410 }
5411
5412 pExpr->iDocid = 0;
5413 pExpr->bEof = 0;
5414 pExpr->bStart = 0;
5415
5416 fts3EvalRestart(pCsr, pExpr->pLeft, pRc);
5417 fts3EvalRestart(pCsr, pExpr->pRight, pRc);
5418 }
5419 }
5420
5421 /*
5422 ** After allocating the Fts3Expr.aMI[] array for each phrase in the
5423 ** expression rooted at pExpr, the cursor iterates through all rows matched
5424 ** by pExpr, calling this function for each row. This function increments
5425 ** the values in Fts3Expr.aMI[] according to the position-list currently
5426 ** found in Fts3Expr.pPhrase->doclist.pList for each of the phrase
5427 ** expression nodes.
5428 */
5429 static void fts3EvalUpdateCounts(Fts3Expr *pExpr){
5430 if( pExpr ){
5431 Fts3Phrase *pPhrase = pExpr->pPhrase;
5432 if( pPhrase && pPhrase->doclist.pList ){
5433 int iCol = 0;
5434 char *p = pPhrase->doclist.pList;
5435
5436 assert( *p );
5437 while( 1 ){
5438 u8 c = 0;
5439 int iCnt = 0;
5440 while( 0xFE & (*p | c) ){
5441 if( (c&0x80)==0 ) iCnt++;
5442 c = *p++ & 0x80;
5443 }
5444
5445 /* aMI[iCol*3 + 1] = Number of occurrences
5446 ** aMI[iCol*3 + 2] = Number of rows containing at least one instance
5447 */
5448 pExpr->aMI[iCol*3 + 1] += iCnt;
5449 pExpr->aMI[iCol*3 + 2] += (iCnt>0);
5450 if( *p==0x00 ) break;
5451 p++;
5452 p += fts3GetVarint32(p, &iCol);
5453 }
5454 }
5455
5456 fts3EvalUpdateCounts(pExpr->pLeft);
5457 fts3EvalUpdateCounts(pExpr->pRight);
5458 }
5459 }
5460
5461 /*
5462 ** Expression pExpr must be of type FTSQUERY_PHRASE.
5463 **
5464 ** If it is not already allocated and populated, this function allocates and
5465 ** populates the Fts3Expr.aMI[] array for expression pExpr. If pExpr is part
5466 ** of a NEAR expression, then it also allocates and populates the same array
5467 ** for all other phrases that are part of the NEAR expression.
5468 **
5469 ** SQLITE_OK is returned if the aMI[] array is successfully allocated and
5470 ** populated. Otherwise, if an error occurs, an SQLite error code is returned.
5471 */
5472 static int fts3EvalGatherStats(
5473 Fts3Cursor *pCsr, /* Cursor object */
5474 Fts3Expr *pExpr /* FTSQUERY_PHRASE expression */
5475 ){
5476 int rc = SQLITE_OK; /* Return code */
5477
5478 assert( pExpr->eType==FTSQUERY_PHRASE );
5479 if( pExpr->aMI==0 ){
5480 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
5481 Fts3Expr *pRoot; /* Root of NEAR expression */
5482 Fts3Expr *p; /* Iterator used for several purposes */
5483
5484 sqlite3_int64 iPrevId = pCsr->iPrevId;
5485 sqlite3_int64 iDocid;
5486 u8 bEof;
5487
5488 /* Find the root of the NEAR expression */
5489 pRoot = pExpr;
5490 while( pRoot->pParent && pRoot->pParent->eType==FTSQUERY_NEAR ){
5491 pRoot = pRoot->pParent;
5492 }
5493 iDocid = pRoot->iDocid;
5494 bEof = pRoot->bEof;
5495 assert( pRoot->bStart );
5496
5497 /* Allocate space for the aMSI[] array of each FTSQUERY_PHRASE node */
5498 for(p=pRoot; p; p=p->pLeft){
5499 Fts3Expr *pE = (p->eType==FTSQUERY_PHRASE?p:p->pRight);
5500 assert( pE->aMI==0 );
5501 pE->aMI = (u32 *)sqlite3_malloc(pTab->nColumn * 3 * sizeof(u32));
5502 if( !pE->aMI ) return SQLITE_NOMEM;
5503 memset(pE->aMI, 0, pTab->nColumn * 3 * sizeof(u32));
5504 }
5505
5506 fts3EvalRestart(pCsr, pRoot, &rc);
5507
5508 while( pCsr->isEof==0 && rc==SQLITE_OK ){
5509
5510 do {
5511 /* Ensure the %_content statement is reset. */
5512 if( pCsr->isRequireSeek==0 ) sqlite3_reset(pCsr->pStmt);
5513 assert( sqlite3_data_count(pCsr->pStmt)==0 );
5514
5515 /* Advance to the next document */
5516 fts3EvalNextRow(pCsr, pRoot, &rc);
5517 pCsr->isEof = pRoot->bEof;
5518 pCsr->isRequireSeek = 1;
5519 pCsr->isMatchinfoNeeded = 1;
5520 pCsr->iPrevId = pRoot->iDocid;
5521 }while( pCsr->isEof==0
5522 && pRoot->eType==FTSQUERY_NEAR
5523 && fts3EvalTestDeferredAndNear(pCsr, &rc)
5524 );
5525
5526 if( rc==SQLITE_OK && pCsr->isEof==0 ){
5527 fts3EvalUpdateCounts(pRoot);
5528 }
5529 }
5530
5531 pCsr->isEof = 0;
5532 pCsr->iPrevId = iPrevId;
5533
5534 if( bEof ){
5535 pRoot->bEof = bEof;
5536 }else{
5537 /* Caution: pRoot may iterate through docids in ascending or descending
5538 ** order. For this reason, even though it seems more defensive, the
5539 ** do loop can not be written:
5540 **
5541 ** do {...} while( pRoot->iDocid<iDocid && rc==SQLITE_OK );
5542 */
5543 fts3EvalRestart(pCsr, pRoot, &rc);
5544 do {
5545 fts3EvalNextRow(pCsr, pRoot, &rc);
5546 assert( pRoot->bEof==0 );
5547 }while( pRoot->iDocid!=iDocid && rc==SQLITE_OK );
5548 fts3EvalTestDeferredAndNear(pCsr, &rc);
5549 }
5550 }
5551 return rc;
5552 }
5553
5554 /*
5555 ** This function is used by the matchinfo() module to query a phrase
5556 ** expression node for the following information:
5557 **
5558 ** 1. The total number of occurrences of the phrase in each column of
5559 ** the FTS table (considering all rows), and
5560 **
5561 ** 2. For each column, the number of rows in the table for which the
5562 ** column contains at least one instance of the phrase.
5563 **
5564 ** If no error occurs, SQLITE_OK is returned and the values for each column
5565 ** written into the array aiOut as follows:
5566 **
5567 ** aiOut[iCol*3 + 1] = Number of occurrences
5568 ** aiOut[iCol*3 + 2] = Number of rows containing at least one instance
5569 **
5570 ** Caveats:
5571 **
5572 ** * If a phrase consists entirely of deferred tokens, then all output
5573 ** values are set to the number of documents in the table. In other
5574 ** words we assume that very common tokens occur exactly once in each
5575 ** column of each row of the table.
5576 **
5577 ** * If a phrase contains some deferred tokens (and some non-deferred
5578 ** tokens), count the potential occurrence identified by considering
5579 ** the non-deferred tokens instead of actual phrase occurrences.
5580 **
5581 ** * If the phrase is part of a NEAR expression, then only phrase instances
5582 ** that meet the NEAR constraint are included in the counts.
5583 */
5584 int sqlite3Fts3EvalPhraseStats(
5585 Fts3Cursor *pCsr, /* FTS cursor handle */
5586 Fts3Expr *pExpr, /* Phrase expression */
5587 u32 *aiOut /* Array to write results into (see above) */
5588 ){
5589 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
5590 int rc = SQLITE_OK;
5591 int iCol;
5592
5593 if( pExpr->bDeferred && pExpr->pParent->eType!=FTSQUERY_NEAR ){
5594 assert( pCsr->nDoc>0 );
5595 for(iCol=0; iCol<pTab->nColumn; iCol++){
5596 aiOut[iCol*3 + 1] = (u32)pCsr->nDoc;
5597 aiOut[iCol*3 + 2] = (u32)pCsr->nDoc;
5598 }
5599 }else{
5600 rc = fts3EvalGatherStats(pCsr, pExpr);
5601 if( rc==SQLITE_OK ){
5602 assert( pExpr->aMI );
5603 for(iCol=0; iCol<pTab->nColumn; iCol++){
5604 aiOut[iCol*3 + 1] = pExpr->aMI[iCol*3 + 1];
5605 aiOut[iCol*3 + 2] = pExpr->aMI[iCol*3 + 2];
5606 }
5607 }
5608 }
5609
5610 return rc;
5611 }
5612
5613 /*
5614 ** The expression pExpr passed as the second argument to this function
5615 ** must be of type FTSQUERY_PHRASE.
5616 **
5617 ** The returned value is either NULL or a pointer to a buffer containing
5618 ** a position-list indicating the occurrences of the phrase in column iCol
5619 ** of the current row.
5620 **
5621 ** More specifically, the returned buffer contains 1 varint for each
5622 ** occurrence of the phrase in the column, stored using the normal (delta+2)
5623 ** compression and is terminated by either an 0x01 or 0x00 byte. For example,
5624 ** if the requested column contains "a b X c d X X" and the position-list
5625 ** for 'X' is requested, the buffer returned may contain:
5626 **
5627 ** 0x04 0x05 0x03 0x01 or 0x04 0x05 0x03 0x00
5628 **
5629 ** This function works regardless of whether or not the phrase is deferred,
5630 ** incremental, or neither.
5631 */
5632 int sqlite3Fts3EvalPhrasePoslist(
5633 Fts3Cursor *pCsr, /* FTS3 cursor object */
5634 Fts3Expr *pExpr, /* Phrase to return doclist for */
5635 int iCol, /* Column to return position list for */
5636 char **ppOut /* OUT: Pointer to position list */
5637 ){
5638 Fts3Phrase *pPhrase = pExpr->pPhrase;
5639 Fts3Table *pTab = (Fts3Table *)pCsr->base.pVtab;
5640 char *pIter;
5641 int iThis;
5642 sqlite3_int64 iDocid;
5643
5644 /* If this phrase is applies specifically to some column other than
5645 ** column iCol, return a NULL pointer. */
5646 *ppOut = 0;
5647 assert( iCol>=0 && iCol<pTab->nColumn );
5648 if( (pPhrase->iColumn<pTab->nColumn && pPhrase->iColumn!=iCol) ){
5649 return SQLITE_OK;
5650 }
5651
5652 iDocid = pExpr->iDocid;
5653 pIter = pPhrase->doclist.pList;
5654 if( iDocid!=pCsr->iPrevId || pExpr->bEof ){
5655 int bDescDoclist = pTab->bDescIdx; /* For DOCID_CMP macro */
5656 int iMul; /* +1 if csr dir matches index dir, else -1 */
5657 int bOr = 0;
5658 u8 bEof = 0;
5659 u8 bTreeEof = 0;
5660 Fts3Expr *p; /* Used to iterate from pExpr to root */
5661 Fts3Expr *pNear; /* Most senior NEAR ancestor (or pExpr) */
5662
5663 /* Check if this phrase descends from an OR expression node. If not,
5664 ** return NULL. Otherwise, the entry that corresponds to docid
5665 ** pCsr->iPrevId may lie earlier in the doclist buffer. Or, if the
5666 ** tree that the node is part of has been marked as EOF, but the node
5667 ** itself is not EOF, then it may point to an earlier entry. */
5668 pNear = pExpr;
5669 for(p=pExpr->pParent; p; p=p->pParent){
5670 if( p->eType==FTSQUERY_OR ) bOr = 1;
5671 if( p->eType==FTSQUERY_NEAR ) pNear = p;
5672 if( p->bEof ) bTreeEof = 1;
5673 }
5674 if( bOr==0 ) return SQLITE_OK;
5675
5676 /* This is the descendent of an OR node. In this case we cannot use
5677 ** an incremental phrase. Load the entire doclist for the phrase
5678 ** into memory in this case. */
5679 if( pPhrase->bIncr ){
5680 int rc = SQLITE_OK;
5681 int bEofSave = pExpr->bEof;
5682 fts3EvalRestart(pCsr, pExpr, &rc);
5683 while( rc==SQLITE_OK && !pExpr->bEof ){
5684 fts3EvalNextRow(pCsr, pExpr, &rc);
5685 if( bEofSave==0 && pExpr->iDocid==iDocid ) break;
5686 }
5687 pIter = pPhrase->doclist.pList;
5688 assert( rc!=SQLITE_OK || pPhrase->bIncr==0 );
5689 if( rc!=SQLITE_OK ) return rc;
5690 }
5691
5692 iMul = ((pCsr->bDesc==bDescDoclist) ? 1 : -1);
5693 while( bTreeEof==1
5694 && pNear->bEof==0
5695 && (DOCID_CMP(pNear->iDocid, pCsr->iPrevId) * iMul)<0
5696 ){
5697 int rc = SQLITE_OK;
5698 fts3EvalNextRow(pCsr, pExpr, &rc);
5699 if( rc!=SQLITE_OK ) return rc;
5700 iDocid = pExpr->iDocid;
5701 pIter = pPhrase->doclist.pList;
5702 }
5703
5704 bEof = (pPhrase->doclist.nAll==0);
5705 assert( bDescDoclist==0 || bDescDoclist==1 );
5706 assert( pCsr->bDesc==0 || pCsr->bDesc==1 );
5707
5708 if( bEof==0 ){
5709 if( pCsr->bDesc==bDescDoclist ){
5710 int dummy;
5711 if( pNear->bEof ){
5712 /* This expression is already at EOF. So position it to point to the
5713 ** last entry in the doclist at pPhrase->doclist.aAll[]. Variable
5714 ** iDocid is already set for this entry, so all that is required is
5715 ** to set pIter to point to the first byte of the last position-list
5716 ** in the doclist.
5717 **
5718 ** It would also be correct to set pIter and iDocid to zero. In
5719 ** this case, the first call to sqltie3Fts4DoclistPrev() below
5720 ** would also move the iterator to point to the last entry in the
5721 ** doclist. However, this is expensive, as to do so it has to
5722 ** iterate through the entire doclist from start to finish (since
5723 ** it does not know the docid for the last entry). */
5724 pIter = &pPhrase->doclist.aAll[pPhrase->doclist.nAll-1];
5725 fts3ReversePoslist(pPhrase->doclist.aAll, &pIter);
5726 }
5727 while( (pIter==0 || DOCID_CMP(iDocid, pCsr->iPrevId)>0 ) && bEof==0 ){
5728 sqlite3Fts3DoclistPrev(
5729 bDescDoclist, pPhrase->doclist.aAll, pPhrase->doclist.nAll,
5730 &pIter, &iDocid, &dummy, &bEof
5731 );
5732 }
5733 }else{
5734 if( pNear->bEof ){
5735 pIter = 0;
5736 iDocid = 0;
5737 }
5738 while( (pIter==0 || DOCID_CMP(iDocid, pCsr->iPrevId)<0 ) && bEof==0 ){
5739 sqlite3Fts3DoclistNext(
5740 bDescDoclist, pPhrase->doclist.aAll, pPhrase->doclist.nAll,
5741 &pIter, &iDocid, &bEof
5742 );
5743 }
5744 }
5745 }
5746
5747 if( bEof || iDocid!=pCsr->iPrevId ) pIter = 0;
5748 }
5749 if( pIter==0 ) return SQLITE_OK;
5750
5751 if( *pIter==0x01 ){
5752 pIter++;
5753 pIter += fts3GetVarint32(pIter, &iThis);
5754 }else{
5755 iThis = 0;
5756 }
5757 while( iThis<iCol ){
5758 fts3ColumnlistCopy(0, &pIter);
5759 if( *pIter==0x00 ) return 0;
5760 pIter++;
5761 pIter += fts3GetVarint32(pIter, &iThis);
5762 }
5763
5764 *ppOut = ((iCol==iThis)?pIter:0);
5765 return SQLITE_OK;
5766 }
5767
5768 /*
5769 ** Free all components of the Fts3Phrase structure that were allocated by
5770 ** the eval module. Specifically, this means to free:
5771 **
5772 ** * the contents of pPhrase->doclist, and
5773 ** * any Fts3MultiSegReader objects held by phrase tokens.
5774 */
5775 void sqlite3Fts3EvalPhraseCleanup(Fts3Phrase *pPhrase){
5776 if( pPhrase ){
5777 int i;
5778 sqlite3_free(pPhrase->doclist.aAll);
5779 fts3EvalInvalidatePoslist(pPhrase);
5780 memset(&pPhrase->doclist, 0, sizeof(Fts3Doclist));
5781 for(i=0; i<pPhrase->nToken; i++){
5782 fts3SegReaderCursorFree(pPhrase->aToken[i].pSegcsr);
5783 pPhrase->aToken[i].pSegcsr = 0;
5784 }
5785 }
5786 }
5787
5788
5789 /*
5790 ** Return SQLITE_CORRUPT_VTAB.
5791 */
5792 #ifdef SQLITE_DEBUG
5793 int sqlite3Fts3Corrupt(){
5794 return SQLITE_CORRUPT_VTAB;
5795 }
5796 #endif
5797
3694 #if !SQLITE_CORE 5798 #if !SQLITE_CORE
3695 int sqlite3_extension_init( 5799 /*
5800 ** Initialize API pointer table, if required.
5801 */
5802 #ifdef _WIN32
5803 __declspec(dllexport)
5804 #endif
5805 int sqlite3_fts3_init(
3696 sqlite3 *db, 5806 sqlite3 *db,
3697 char **pzErrMsg, 5807 char **pzErrMsg,
3698 const sqlite3_api_routines *pApi 5808 const sqlite3_api_routines *pApi
3699 ){ 5809 ){
3700 SQLITE_EXTENSION_INIT2(pApi) 5810 SQLITE_EXTENSION_INIT2(pApi)
3701 return sqlite3Fts3Init(db); 5811 return sqlite3Fts3Init(db);
3702 } 5812 }
3703 #endif 5813 #endif
3704 5814
3705 #endif 5815 #endif
OLDNEW

Powered by Google App Engine
This is Rietveld 408576698